Spark Noir Creative reads the marketing press each weekday morning and keeps the
stories that change how a business should show up: AI search and the answer engines, new channels
and tools, brand positioning, and the launches worth studying. Every headline links to its
original reporting, and nothing here is Spark Noir Creative's own claim.
Spark Noir Creative · 303 stories across 26 days · Updated 2026-10-14 · Reviewed September 2026
10 stories
[ AI search ]
Google's newest model now runs AI Mode, and this time the answers carry links
On October 13, Google made Gemini 4 Argon available in AI Mode for AI Pro and Ultra subscribers worldwide, describing it as better at complex, multi-step tasks and at following instructions across a long conversation. Search Engine Roundtable notes that Argon's AI Mode responses include links, unlike the earlier Gemini 3.8 Flash release, which shipped without them. No traffic or citation figures were published. (Search Engine Roundtable)
Google is testing the AI Mode sidebar on the regular search results page
Spotted October 14: some users see AI Mode's left sidebar (chat history, a "new AI Mode chat" button, settings) on the standard results page, where before it only appeared after clicking into AI Mode. Barry Schwartz could not reproduce it and calls it a test. Marie Haynes reports her Gemini app conversations already show up in the AI Mode sidebar, which points to Gemini and Search sharing one history. (Search Engine Roundtable)
Google is hiring to turn Discover from a feed into a "concierge"
A Google posting for a Principal UX role on Search Discover (base pay $275,000 to $382,000) says the goal is "a proactive, hyper-personalized concierge experience" that recommends information, products, services and next steps toward a user's goals. A related Staff ML Engineer posting covers AI-generated content in the Discover feed. Google has not commented, and a job post is direction, not a launch. First flagged October 11 by Damien Andell. (PPC Land, Search Engine Journal)
Facebook now lets people type what they want in their feed, and LLMs sit inside the ranking
Meta announced on October 13 a "Your algorithm" space in the three-dot menu where users add topics they want, written follow-up questions after "Interested" or "Not interested," and LLMs in the ranking pipeline so a typed request can change recommendations in the same session. Meta gave no rollout schedule or figures for the feature; earlier this year it credited ranking work with a 10% lift in Instagram Reels time spent. It has not said whether typed interests feed ad targeting. (PPC Land)
Meta credits one ads foundation model with an 18.9% conversion lift, and none of it is a setting you can touch
Meta's engineering blog (October 13) says its Generative Ads Model (GEM) lifted conversions 18.9% and click-through rate 12% across surfaces, with 5.8% on Facebook and 4.1% on Instagram for "targeted users." The figures are Meta's own, without a stated baseline or test design, and they do not reconcile with the 15.7% Facebook figure Meta gave on its July earnings call. The gains reach advertisers only through Advantage+ and automated delivery. (PPC Land)
Liquid Death runs a "pretty sizeable" media budget with two internal people
At Advertising Week (reported October 7), Liquid Death's chief media officer described working inside agency Power Digital's AI stack (an intelligence layer, an operating system and a creative layer), moving retail media optimization from weekly to daily, and automating any task that takes more than five minutes. The agency built 17 versions of its operating system in three weeks. No performance numbers were shared. (Marketing Dive)
Physical channel: no new mail or out-of-home study this week, but 93% of shoppers still plan a store trip and 61% will shop with AI
ICSC's Holiday Intentions Survey (released October 6; 1,003 U.S. adults, fielded September 21 to 23) found 93% expect to visit a physical store this season, 97% of Gen Z. 34% of store shoppers call the trip a social outing, 48% among Gen Z. 61% plan to use AI tools, up from 47% in 2025, mostly to compare prices (29%). 58% plan to try a new brand and 52% a new retailer. (ICSC)
Lippincott and Bloomberg Media: "the pursuit of internal trust can undermine external trust"
Research presented at Advertising Week (reported October 7) found CMOs gain credibility inside their companies by leaning into measurable performance marketing, at the cost of cultural relevance and brand love. 41% of CMOs say they are culturally tuned in but their organizations lag, and 81% of those doing cultural marketing would do more without obstacles. Only 11% rate their organization's ability to adopt new marketing technology as excellent. (Marketing Dive)
Food brands are launching handbags, and the $89 Betty Crocker bag sold out in a day
Chipotle is giving away a 3-foot-wide, 170-liter "Big Ass Chipotle Bag" through a multiweek sweepstakes for Rewards members, built to promote its digital-only Family Meals. Unreal Snacks made a $425 silver clutch with designer Rachel Comey that looks like its chocolate wrapper, shown at New York Fashion Week. Betty Crocker's $89 cake-slice bags sold out within a day on August 25, and a second drop drew thousands queued at once. (Marketing Dive)
Asics made a 10-minute spoof nature documentary narrated by Sigourney Weaver for World Mental Health Day
Adweek's Ads of the Week (around October 9) featured Asics' satirical nature film narrated by Weaver, released ahead of World Mental Health Day on October 10, from Golin Ketchum London. The same roundup includes Twix bringing back "Zombie Claus" for a second year of "Hallowmas," from adam&eve\TBWA, and an Ikea dating game show inspired by a wardrobe. (Adweek, Ad Age)
The one-line read. Being talked about and being used as a source are now measurably different things. Semrush's new AI Visibility Index found that fewer than half of the brands AI assistants name most often are also among the sites they cite, and one UK grocer ranked 8th for ChatGPT mentions but 606th for citations. At the same time, the tools businesses use to check their rankings are being fed bad data: one tracker says Google returned spurious results on 11% of its daily checks. Shopping agents are picking winners on their own rules, with Meta's Muse reportedly told to put well-known merchants in its top five. The brands that stood out this week did it with specifics, not adjectives: Omaha Steaks dropped "gourmet" for sourcing and aging details and saw conversions rise 46%, and Espolòn refunds half of a Halloween surge-priced ride. The condition for a business that wants to be found: name recognition gets a brand into the answer, but only pages worth citing get it the link, and the dashboards measuring either need a second look this month.
[ AI search ]
Brands AI names most often are usually not the ones it cites: Asda ranks 8th in ChatGPT mentions and 606th in citations
Semrush's H2 2026 AI Visibility Index, published October 13, measured ChatGPT, Gemini, Google AI Mode and AI Overviews monthly from January to July 2026 on fixed pools of 25 million US and 2.5 million UK prompts. Out of the top 100 most-mentioned brands, only 41 to 47 in the US were also among the top 100 cited domains (36.7 to 45.3 in the UK), and for ChatGPT alone the US overlap fell to 33 to 36. Semrush calls the gap a "one-way door": 8 to 13 US domains a month were highly mentioned but rarely cited, while the reverse almost never happened. ChatGPT and AI Overviews shared the fewest sources of any pair, about 43 of 100 in the US.
Follow-up: rank trackers are being fed fake Google results, so September and October ranking data is suspect
DemandSphere says Google has started serving "spurious results" to suspected bot traffic, affecting about 11% of its daily rank checks, and Nozzle measured 8.0% of its pages as poisoned; Google has not confirmed either. A blocked request leaves a visible gap, but a poisoned one looks normal and reports a confident wrong ranking. DemandSphere's own branded AI Overview presence series, which jumped from 26.12% on September 1 to 82.06% on September 29, came through the same pipeline under heavy blocking and should be read with caution. Google's September spam update also ran September 24 to October 8, so real volatility and data noise overlap.
AI Overviews now answer typos and searches with no results
Search Engine Roundtable reported on October 12 that AI Overviews appear for misspelled queries, random keyboard input and searches where Google finds no matching pages. For one made-up query the Overview replied that the message "might have been a typo or random keyboard input." Google has also been labeling old utilities, such as a character counter, as AI Overviews.
Follow-up: Meta's Muse shopping agent was reportedly told to put well-known merchants in its top five
Meta's Muse launched in the US on September 8 and passed an estimated 5 million installs in 22 days. An instruction file that AI visibility vendor Profound says it extracted on September 25 makes ranking well-known merchants in the top five picks "imperative," caps open-web search at three stores and ten products, discards listings priced far outside a category's normal range, and puts a named brand's own store first. In Profound's 50-prompt test, Meta's catalog supplied about 95% of candidates but won only about a third of the final slots. Meta has not commented and has not defined "well-known."
Meta's ad assistant can now change budgets and targeting from a chat prompt, and takes your brand voice file
At Advertising Week New York on October 8, Meta said its business assistant can now generate creative, update targeting and change an ad's budget from conversational prompts. It is testing an import tool so brands can upload strategy and brand-voice guidelines, plus scheduled recurring performance reports and alerts. Meta also widened access to Ads Creative Studio and expanded its Ads MCP server so outside AI tools can build and manage campaigns. Meta gave no adoption or performance figures.
Follow-up: Local Services Ads are becoming a Performance Max campaign, and manual cost-per-lead bidding goes away
Google is moving Local Services Ads into Google Ads as a pay-per-lead Performance Max campaign limited to Search and Maps. The first phase started in August for US home-service categories such as plumbing, HVAC and roofing; service-area businesses migrate late 2026 and the remaining categories in 2027 (real estate's date is not stated). Manual maximum cost-per-lead bidding is removed, weekly budgets become daily, historical reports do not carry over, and the old dashboard closes 14 days after the notice email. Unanswered business-hours calls longer than 20 seconds have been billable since October 1. Separately, lead-response tool LSArespond says it passed 2,000 users and 115,000 leads processed in August.
Physical channel: no new consumer mail or sign study this week, but WARC says digital out-of-home grows 13.7% this year
WARC's global ad forecast, reported October 12, puts 2026 digital out-of-home spend up 13.7% to $21.7 billion, ahead of total global ad growth of 11.9% to $1.34 trillion. Social media is the fastest-growing channel at 21.3% to $394.6 billion, and Alphabet, Amazon and Meta take 59.7% of spend outside China. No new nationwide direct mail, print or yard-sign consumer study surfaced in the last seven days.
Omaha Steaks dropped "gourmet" for specifics, and conversions rose 46% in two weeks
Omaha Steaks launched its "No Bull" platform in September, moving away from vague words like "gourmet" and "quality" and from leading with discounts. The ads instead say what quality means: US sourcing, aging and hand-cutting, and the 30-second spot opens by addressing the brand's age head-on. The company reported conversion rates up 46% and social mentions up 130% about two weeks after launch. It backed the message with operations: delivery time fell from more than six days to about two, and direct mail stays a core channel for customers who prefer something tangible.
Thorne found customers cared less about its science than it thought, and a consumer-first platform preceded a $3.8 billion sale
Thorne's growth chief told Marketing Dive (September 29) that research with Redscout, using digital diaries, a nationally representative survey and small-group interviews, showed consumers cared less about the supplement maker's scientific rigor than the company expected. Thorne moved from a Thorne-first message to a consumer-first one: "Thorne rewards curiosity with excellence," with the science as support rather than the lead. The "Now I Know" campaign followed in January; direct-to-consumer sales grew 63% in 2025, and P&G agreed to buy the company for $3.8 billion in August.
Espolòn pays back half of a Halloween surge-priced ride, up to $25
Campari's Espolòn Tequila launched "Fare Share," refunding 50% of an Uber or Lyft fare, up to $25, for rides taken during qualifying hours through November 1. The brand cites big-city ride costs reaching eight to nine times normal on Halloween last year. It runs on earned media, creators, bar and restaurant signage and in-person activations at Halloween and Día de los Muertos events in Los Angeles, New York, Dallas, Houston and Austin, with a landing page for receipt submission.
Reese's is crowning the "Trick-or-Treatiest Street in America," with Zillow as the partner
Reese's surveyed 15,000 Americans to pick 150 streets, and a first-ever "Halloween Council" of creators and celebrities is judging them on social; the winner gets a block party from Reese's and Zillow, announced October 21. A microsite maps all 150 streets, and a two-part social series documents the judging. The National Retail Federation expects $13.5 billion in Halloween spending, $4.1 billion of it on candy.
The one-line read. Search engines and AI crawlers are tightening what they will show and what they will read. Google quietly stopped showing review stars on healthcare pages even when the markup is correct, it is testing an "add to calendar" link inside AI Overviews, and today it begins pulling promotions off advertisers' own pages into ads by default. On the other side of the fence, sites on Cloudflare's network turned away 13.47% of AI bot requests in the third quarter, more than double a year earlier, which means some businesses are blocking the very engines they want to be cited by. The brands that stood out this week did it by going backward to go forward: Southwest's new marketing chief rebuilt its campaign from the airline's archives and paid people in points for screenshots of their notifications, and MorningStar Farms put "Goodness Since 1974" on the box. The condition for a business that wants to be found: the free visual extras in search are shrinking, so the plain facts on the page, and whether an AI is even allowed to read them, now carry more of the load.
[ AI search ]
Google stopped showing review stars on healthcare pages, even with correct schema
Search Engine Roundtable reported on October 7 that Google no longer shows review rich results for healthcare web pages that carry review structured data. Schema App, which first spotted it, said its healthcare clients' review snippet visibility began falling in late May, with a second drop in early August, and none of its markup fixes (rewiring aggregateRating, multi-typing entities as Product) brought the stars back. Google has not commented. It follows Google's 2019 rule that removed self-serving review stars from LocalBusiness and Organization pages.
AI crawlers were turned away 13.47% of the time in Q3, more than double a year ago
A release from Zen Media on October 9, built on TechnologyChecker.io's read of Cloudflare data, says the share of AI bot requests denied on Cloudflare's network rose to 13.47% in Q3 2026 from 5.98% in Q3 2025 (Q3 figures preliminary). Search and user-triggered AI requests, the kind that fetch a page to answer a live question, rose from 6.51% to 11.92% of AI crawl activity between Q2 and Q3. In a snapshot of 4,242 robots.txt files on September 28, GPTBot was blocked 1.60 times for every time it was allowed, and OAI-SearchBot 0.91 times. Zen Media sells an AI visibility audit, so read the framing as a pitch, but the numbers are specific.
Google is testing "add to calendar" inside AI Overviews
On October 7, Search Engine Roundtable showed screenshots of AI Overviews with an "add to calendar" link next to event dates, first spotted by Khushal Bherwani and then Radu Oncescu. The link disappeared on refresh and Barry Schwartz could not reproduce it, so it is an early test.
Follow-up: Google Ads automated promotions switch on today
As covered October 5, Google Ads begins pulling offers from advertisers' landing pages into promotion assets starting today, October 12, on Search and Performance Max campaigns that have location assets linked and no promotion assets of their own. It is on by default; the off switch sits in account-level automated asset settings. Each sourced promotion must be re-verified within 24 hours, and Google has not defined what counts as "high quality."
A retailer that switched to "reject all" cookies says it got 6% more search revenue on 5% less spend
Corvidae AI said on October 9 that RS Group, after moving to a reject-all cookie consent model, fed reconstructed customer-journey data into its automated bidding and saw 6% more paid search revenue on 5% less ad spend. A four-month pilot in Italy lifted return on ad spend 12%, and the reconstructed journeys ran 46% longer in the UK and 68% longer in the EU than the ad platforms reported. The release gives no control group, attribution window or baseline, so these are vendor-reported results.
Martech is plugging straight into the AI assistants: Nectar opened an MCP server, Hightouch bought Strata
On October 9, Nectar, a social and creator data platform, launched an MCP server that lets ChatGPT, Claude and Gemini read a brand's comments, mentions, creator activity and competitor tracking through 16 read-only tools (Copilot is listed as coming soon). The same day, Hightouch acquired Strata, a 2024 startup whose AI builds and maintains no-code integrations between marketing tools; terms were not disclosed. Nectar has raised $40.6 million and names e.l.f. Beauty, Sweetgreen and Liquid Death as customers.
HeyGen: 71.6% of small business owners recorded a video for their business and never posted it
HeyGen launched its own voice model on October 9, free inside its platform, with a $99 per month Professional Voice Clone trained on 30 minutes to 3 hours of the owner's speech and requiring the owner's explicit consent. Alongside it, HeyGen said a survey of more than 1,000 small business owners in its user base found 71.6% had recorded a business video and then decided not to post it. The company reports more than 40 million users and over $200 million in annual recurring revenue.
Physical channel: no new mail or out-of-home study this week, but MLB bought full newspaper pages to start the playoffs
No new direct mail, print or out-of-home consumer study surfaced in the last seven days; the most recent remain Lob's September 15 direct mail report and OAAA's September experiential research, both covered earlier. The clearest physical-channel play came from Major League Baseball, which ran full-page newspaper ads reading "October Starts Today" in postseason markets, including The New York Times and the LA Times, at the end of September, under the platform "No Time Like October" from agency The Team. Adweek reports Fox and TNT Sports saw year-over-year playoff ratings gains, but gives no figures tying them to the ads.
Southwest's new marketing chief went to the archives and merged marketing with digital
The Drum interviewed Sabrina Callahan on October 6. She joined Southwest in February 2026 as its first Chief Digital and Marketing Officer, combining marketing and digital for the first time, and spent much of her first eight months in the airline's archives, research and founders' philosophy. She describes the old brand as "a little bit rebel with a cause," playful and warm, and wants that back; she added Foundation Media Partners on a multi-year deal for entertainment and talent tie-ins. Her stated priorities are consideration, bookings and repeat customers, because "awareness is not the main challenge." Her aim is that customers never feel they are "talking to an org chart."
MorningStar Farms put "Goodness Since 1974" on the box while newer rivals struggle
MorningStar Farms, owned by Kellanova, began rolling out a new identity on September 28, designed by Design Bridge and Partners with Kellanova's in-house teams. It is built on the platform "Where Goodness Grows," with deep farm greens, warm 1970s-style lettering, woodcut-style illustrations and "Goodness Since 1974" emblems. Fast Company frames it as using the brand's 52 years to tell shoppers that plant-based food is not a fad, at a moment when Beyond and Impossible are reporting declining revenue. The new packaging launched first on new items and is moving across the range.
Southwest's "Enter Airplane Mode" pays people points for a screenshot of their notifications
Southwest launched "Enter Airplane Mode" on October 5, created by longtime agency GSD&M, directed by Will Speck and Josh Gordon and narrated by Adam Scott. It frames a booking as an escape from burnout and cites a Southwest-sponsored survey in which 76% of respondents said they are burned out. The mechanic: in "Notification Vacation," Rapid Rewards members upload a screenshot of their phone's notifications to earn points, and a Burnout Quiz also pays points, with more than 185 million points available in total. Rapid Rewards enrollment grew 35% year over year in Q2.
Hidden Valley made its first seasonal product: black ranch in a glow-in-the-dark bottle
Clorox's Hidden Valley released Spooky Ranch on September 24, its first seasonal product in the brand's history: Original Ranch flavor colored black with fruit juice and cocoa, in a 12-ounce bottle that glows in the dark. It is a Walmart exclusive at $3.97, sold only for Halloween while supplies last. No sales figures have been released.
The one-line read. Trust is the scarce input this week, and the data shows where it is leaking. SOCi's Local Discovery Index found that 52% of U.S. consumers used an AI tool to find a local business in the last 30 days, up from 9% a year earlier, yet 81% check the answer somewhere else before they call and 67% have been given wrong information. Simon-Kucher found only 22% of shoppers believe advertised discounts fully reflect real savings, down from 46% last year, and Omnicom found 65% of consumers avoid ads to some degree. Meanwhile the machinery keeps moving toward the transaction: Google is testing one-time promo codes inside AI Mode, Newegg now checks out inside Gemini and AI Mode, and Meta, MessageGears and Amperity shipped agents that build campaigns but leave the final send to a person. The brands that stood out made themselves simpler to recognize: Intuit pulled TurboTax, QuickBooks, Credit Karma and Mailchimp under one identity, Lucky Energy stripped its cans to black and white, and Wet Ones turned a single cheeky image into a citywide out-of-home takeover that a rival answered within days. The condition for a business that wants to be found: being named by an AI is only the first step, because the customer verifies next, and the business with consistent reviews, plain prices and a recognizable identity is the one that survives the check.
[ AI search ]
52% of consumers used AI to find a local business last month, but 81% verify before they call
SOCi's 2026 Local Discovery Index, a weighted survey of more than 1,000 U.S. consumers, found the share who used an AI tool to find a local business in the past 30 days jumped from 9% in 2025 to 52% in 2026. After an AI recommendation, 81% take a verification step first: 33% check reviews, 20% check the business's social media, 16% search independently and 12% check several sources; only 19% contact the business directly. Among AI users, 67% have received wrong information about a local business at least once and 30% said it caused a real inconvenience. Reviews still decide: 99% read them before a first visit, and 72% are more likely to choose a business that responds to reviews, up from 65% in 2025. The index was published September 24 and drew wider coverage this week.
Google is testing one-time promo codes inside AI Mode, and AI Mode is reportedly showing three bottom ads instead of two
A beta "Direct offer" asset appeared in Google Ads on October 7, sitting beside Promotions, Messages and Lead forms. It only serves on campaigns using AI Max or text customization. Advertisers give a name, URL and short description; Google's AI decides when the offer is relevant and can write the offer text itself, then shows a "Claim one-time code" option under a sponsored result in AI Mode. Advertisers can upload single-use codes by CSV, set several incentives and let Google pick, and cap daily claims by value or count. Separately, Search Engine Watch reported AI Mode now shows three ads at the bottom of answers rather than two.
Newegg now sells inside Gemini and Google AI Mode, paid with Google Pay
Newegg announced on October 8 that U.S. shoppers can buy eligible products directly in the Gemini app and in Google Search, including AI Mode, paying with Google Pay. It runs on Google's Universal Commerce Protocol (UCP), and Newegg stays the merchant of record. The release gives no count of eligible products and no order data.
Meta opened Customer Lifecycle Strategy to all advertisers and made static-to-video generation generally available
At Advertising Week New York on October 6, Meta said Customer Lifecycle Strategy is "now available to all advertisers" and that video generated from static images in Advantage+ creative is generally available. Ads Creative Studio is rolling out to more advertisers, and the Meta AI business assistant is starting to test agentic actions (campaign setup, targeting and budget changes) with no date given. AI-enabled audience discovery in plain language is due "by the end of the year." Meta says assistant users typically see 5% more conversions at the same cost per conversion, without explaining the method.
Agents now build the campaign, but people still press send, and most consumers are not ready either
On October 8, MessageGears released an MCP server that lets marketers build audiences and campaigns from Claude or ChatGPT, but agents cannot launch, schedule or delete email, SMS or push sends; a person launches from the MessageGears interface. Amperity's new agent, Pér, prepares audiences and actions but users "review, refine and approve the next step before Pér acts." The same day, Channel V Media published a Prosper Insights survey of 7,675 U.S. adults: 40.1% would let an AI agent handle at least one everyday task, but only 7.7% use agentic AI today, 17.6% think agents are a good idea, and 69.1% prefer a person for online shopping.
Physical-channel check: no new direct mail or out-of-home study this week, but 65% of consumers say they avoid ads
No new direct mail, print or out-of-home consumer study was published in the last seven days; Lob's September 15 report was covered earlier. The closest nationwide consumer study is Omnicom Media Intelligence's, released October 6: in a survey of 1,136 U.S. consumers, 65% avoid ads to some degree (25% very or extremely actively), 41% pay for subscriptions to avoid ads, and 55% increasingly use apps and AI instead of websites. 79% say they are more likely to remember a brand from sponsored content they enjoy than from a traditional commercial, and 84% say brands should try new ways to connect.
Only 22% of shoppers believe advertised discounts fully reflect real savings, down from 46%
Simon-Kucher's eighth annual Holiday Shopping Study, released October 8 and based on 1,501 U.S. consumers surveyed August 6 to 20, found just 22% believe advertised discounts fully reflect true savings, down from 46% in 2025. PPC Land notes that another 45% say discounts "mostly" do, so the combined figure fell more gently, from 71% to 67%. Planned holiday spending drops 5.5% to $1,018 per household, 77% say inflation changed how they shop, and 55% plan to use AI to find deals, though only 16% of those would let AI complete a purchase.
Intuit put TurboTax, QuickBooks, Credit Karma and Mailchimp under one identity, because customers did not connect them
Creative Review reported on September 30 that JKR and Intuit's in-house team built "One Intuit": a new wordmark and shortened "i" mark used across app icons and campaigns, a custom typeface called Tally, a lead color called Superblue, and "Intuit Intelligence" at the center, telling customers whether they are dealing with AI, a human expert or both. JKR's Michael Ciancio said people often do not link the value of Intuit's products to the parent company, and research showed that when they do, trust rises and they adopt more products. No metrics were published.
Lucky Energy went black and white and named the gap it sits in, without changing the product
At the NACS Show on October 7, Lucky Energy unveiled a Dayjob-designed identity: black-and-white cans with a bolder wordmark where the only color is the flavor graphic, and a stated position between performance energy drinks and newer wellness brands. The formula, caffeine, can size and price are unchanged. The new mission line is "Every day is a miracle." Lucky is targeting 17,000 retail locations by the end of 2026, including 218 H-E-B stores and 380 Walmart Fuel Centers.
Wet Ones took over New York with one cheeky image, and Goodwipes answered with a billboard next door
Wet Ones and BBH USA launched "Just For Hands" on October 5: fingers posed to look like a bottom, telling people the wipes are for hands only. The NYC buy covers 100 LinkNYC kiosks, 60 transit shelters, 20 digital newsstands, a five-screen Times Square billboard, 40 wild-posting sites, two wrapped buses and six "Take a Wet One" sampling boards, with about 4,000 samples handed out. Fans can post their own "hand butt" photo to appear on the Times Square screen, and the brand says subway placements were refused as too risqué. Within days Goodwipes, which makes wipes that are meant for that use, placed a pink out-of-home ad next to a Wet Ones placement, and Ad Age named Goodwipes a winner of the week.
The one-line read. The answer box is turning into the tool. OpenAI put GPT-6 into ChatGPT with an "Intelligent UI" that builds calculators, forms and charts inside the conversation and starts search answers 44% sooner, which puts every simple calculator page on the web in direct competition with the chat itself. Google finished a two-week spam update while practitioners report a jump in manual actions against long-standing businesses, and it is testing a WhatsApp button in local panels. On the buying side, 17.5% of 36,289 Meta ad batches at one bulk tool were placed by AI agents in September, and Google opened Conversion Lift tests to any advertiser with 1,000 conversions and $5,000. The brand stories ran the other way: Les Binet found consumer brands spend twice as much on price promotions as on paid media, Tesco reported its highest customer satisfaction score on a brand platform built around small surprises, and Adweek argued that distrust of AI content is pushing advertisers back into the physical world. The condition for a business that wants to be found: a page worth citing beats a page worth clicking, real people beat production value, and the offline world is gaining credibility as the online one loses it.
[ AI search ]
ChatGPT moved to GPT-6 with an "Intelligent UI" that builds calculators and forms inside the answer
OpenAI is rolling out GPT-6 to all ChatGPT users, paid plans first and Free and Go the following day, replacing the "Latest" model option. The new Intelligent UI lets answers mix text with buttons, forms and charts, and users can ask ChatGPT to build simple tools such as a savings calculator or a bill splitter inside the conversation. For web search questions, GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant, and OpenAI says it is better at deciding when to search. The announcement does not say how sources will be shown when an answer is a tool rather than text.
Google's September spam update finished after nearly 14 days, and SEOs report a jump in manual actions against real businesses
The September 2026 spam update started September 24 and finished October 8 at 4:37 am ET, about 13 days and 17 hours, far longer than the roughly two-day spam updates before it. It applies globally, does not target link spam, and Google warns recovery "can take many months." Separately, Marie Haynes said her manual action requests went from "essentially no requests" to "several per week," mostly for thin content, pure spam and scaled content, and mostly on long-standing legitimate businesses; Search Engine Roundtable found 12 recent forum threads on the same theme.
Google is testing a WhatsApp button in local business panels
Google is showing a WhatsApp button in local panel results, either in place of the chat option or alongside call, directions, website and share, so searchers can message a business directly. Barry Schwartz reproduced it and said it "might be rolled out fully." Google earlier asked businesses to add WhatsApp numbers to their Business Profiles, and months later began adding numbers itself for some businesses.
17.5% of 36,289 Meta ad batches at one bulk tool were placed by AI agents in September
Ads Uploader, a bulk Meta ad submission tool, reported that 17.5% of the 36,289 batches it handled from September 1 to October 1 came from AI agents such as Claude or ChatGPT connected through its MCP server or command line, about 212 agent batches a day. The batches created 286,137 Meta ads, 84% went into existing campaigns, and 6.6% carried Meta's AI-generated content label. PPC Land notes the figures are one vendor, one month and one platform, unaudited, and the "agent" label describes the submission channel, not how much a human reviewed.
Google Ads opened Conversion Lift tests to advertisers without a Google rep: 1,000 conversions and $5,000 minimum
Google updated its help documentation so advertisers can set up Conversion Lift measurement themselves; Search and Performance Max studies previously required a Google representative. Eligibility is at least 1,000 observed conversions, a $5,000 minimum campaign budget and a compatible conversion action, across Display, Search, Video, Demand Gen, App and Performance Max. A campaign can sit in only one lift study at a time.
Local Services Ads follow-up: an unanswered call is now billed if the caller holds on for 20 seconds
This is a follow-up to the October 5 report on missed calls becoming billable. Google expanded its "how leads work" document for Local Services Ads: an unanswered call during business hours can be charged if the caller stays connected more than 20 seconds, Google can keep evaluating a lead for 30 days and issue a credit later, and repeat contacts from the same customer are generally not charged again within 15 days. The long list of credit-exclusion examples was removed in favor of Google's automated lead-quality assessment.
Liquid Death is building a marketing organization where two people run a sizeable media budget
Speaking at Advertising Week New York on October 5, Liquid Death's chief media and digital commerce officer Benoit Vatere said the goal is to run "a pretty sizeable budget" with "only two people internally," and "I don't want to see any spreadsheets anymore." Working with Power Digital's AI system, Liquid Death has shipped 17 iterations of it in three weeks, and retail media optimization moved from weekly or twice-weekly reviews to daily. The rule of thumb offered: automate any task that takes more than five minutes.
Physical channel: distrust of AI content is pushing advertisers back into the physical world
Adweek's Mark Stenberg argued this week that answer engines and agents are cutting brands off from the web, while generative AI erodes trust online: 87% of US consumers in a 2024 Adobe study said AI has made it harder to tell fact from fiction. The money is following: US out-of-home revenue rose 10.7% in Q2 to a record $3.16 billion, its first quarter over $3 billion, with digital out-of-home up 18.5% and making up 38.4% of the category, and 73% of the 100 largest OOH advertisers spending more year over year. Droga5's former executive creative director Mariano Jeger left for billboard company Outfront Media, citing trust in online content.
Les Binet: consumer brands spend twice as much on price promotions as on paid media, and 75% cannot say which promotions make money
New research by Les Binet and Kantar's Dom Boyd, surveying 250 marketing and insight leaders at UK consumer goods firms through Censuswide, found price promotions are now the largest single line of marketing spend, about twice paid media. 75% of respondents cannot estimate what share of their promotions are profitable, 80% think many promotional buyers would have bought anyway, 75% believe repeated promotions raise price sensitivity, and 81% judge promotions over windows shorter than 13 weeks. Binet compared promotions to a "class A drug" and proposed cutting the money-losing ones and reinvesting in brand advertising.
Tesco's "Need anything from Tesco" platform is built on small surprises, and UK NPS hit its highest ever at 33
Tesco CEO Ken Murphy told Marketing Week the brand is "designed to surprise" customers: the "Need anything from Tesco" platform, launched in March, aims to show up in small moments when Tesco meets a need, paired with a sharpened value position as the UK's "cheapest full-line grocer." In interim results, UK NPS rose 3 points to 33, its highest ever, UK market share reached 27.8% (up 113 basis points over four years), Tesco Finest sales rose 8.9% in the UK, and full-year profit guidance was raised to £3.15 billion to £3.30 billion.
Chime's growth chief: "Brand or performance? That's the wrong question," and scrappy testimonials beat cinematic ads
At Brandweek 2026, Chime SVP of Growth Nick Fairbairn said brand lift and direct-response accountability belong in the same system, and that he measures TV with performance rigor (traffic spikes, enrollments, real-time signals), as he did to justify a Super Bowl buy at Dollar Shave Club. Chime has seen lower-production testimonial creative lift consideration and brand-love measures, sometimes beating cinematic work: "production value should not be confused with brand impact." He expects creators to become "a significant share of every channel."
JPMorganChase reached women's sports fans at the Final Four in Phoenix without paying for league rights
Chase has no women's league sponsorship, so it worked with media and commerce company Togethxr, which sets up pop-up venues near major women's sports events. At the Women's Final Four in Phoenix this spring, Chase ran a pop-up shop with Golden State Valkyries guard Tiffany Hayes and Round21 where cardholders could access product; at WNBA All-Star Weekend in Chicago it hosted a live podcast taping with cardholder presale tickets. Chase found only a 5% to 6% overlap between its WNBA and NBA card transactions, meaning a separate audience. Deloitte projects women's elite sports at $3 billion or more in 2026, up 340% since 2022.
Bacardí brought back the Crypt Keeper after 30 years as the DJ host of AMC's 30th FearFest
Bacardí and AMC Global Media revived the Crypt Keeper, voiced again by original actor John Kassir and redesigned and directed by horror effects artist Greg Nicotero, as a cocktail-loving DJ hosting five vignettes across AMC's linear TV and social during FearFest's 650 hours of October horror programming. The character carries Bacardí's "Better With a Beat" music platform and a seasonal "Spookito" cocktail, and the stunt coincided with all seven seasons of *Tales From the Crypt* streaming on AMC+ and Shudder for the first time. The team's brief was to keep it from feeling like an ad.
The one-line read. AI assistants are deciding who is "the best" local business, and they are not reading the same sources. A new study of 1,650 answers found ChatGPT named only 3.4% of local firms with fewer than 25 reviews, cited Yelp in none of its answers and leaned on the Better Business Bureau and licensing boards, while Claude cited Yelp in 202 of 218. ChatGPT is also testing "best for beginners" style labels on products, and Google started telling site owners how many readers chose them as a Preferred Source. Ad platforms kept turning into agents: TikTok opened a 400,000-app ad network with agentic lead-gen tools and a connector for Claude and Perplexity. On the brand side, Jaguar finally showed the car behind its rebrand at a price roughly double its old average, and Starbucks reported four straight quarters of growth from a positioning built on being human. The condition for any business that wants to be found: volume of real reviews, presence on the directories each model trusts, and a clear, human reason to choose it.
[ AI search ]
ChatGPT named 3.4% of local businesses with under 25 reviews, and it does not read Yelp at all
8ight.AI asked ChatGPT, Claude and Gemini "Who is the best {trade} in {city}?" across 39 trades and 49 locations (32 US cities plus 17 Toronto-area towns), matching 2,976 local firms against Google Maps, with data collected September 10 to 30 and published October 5. ChatGPT named 3.4% of firms with fewer than 25 reviews, 10.0% at 25 to 99, 13.9% at 100 to 249 and 22.2% at 1,000 or more; star rating showed no correlation with being named. The sources split sharply: Claude cited Yelp in 202 of 218 answers (92.7%), ChatGPT in none of 1,412, instead citing the Better Business Bureau (38.7%), government and licensing sites (25.1%) and Expertise.com (16.3%).
Google is emailing site owners how many readers picked them as a Preferred Source
Search Console began sending some site owners a single number on October 5 and 6: "people are choosing [domain] as a preferred source," with a link to a seven-question survey on whether the figure belongs in Search Console itself. Preferred Sources lets readers choose outlets they want to see more often in Top Stories, AI Overviews and AI Mode; selections grew from 90,000 unique sources in December 2025 to 345,000 in May and more than 600,000 by August 20. Google says readers click about twice as often on sources they chose, without publishing the baseline.
ChatGPT is testing "best for" labels on product results
On October 6, Brodie Clark spotted ChatGPT adding colored labels to product recommendations: "Best all-rounder" in green, with category labels such as "Best for beginners" and "Best cushioning" in gray. It expands ChatGPT's earlier product labeling and is aimed at making the differences between options obvious at a glance. It arrives in the same week ChatGPT ads already carry product feeds into answers.
Google Maps now bans pricing, limited-time offers and calls to action in business listings
Google updated its Maps contribution policy on October 6. Business listings may not contain "time-bound promotions, pricing, or direct calls-to-action," and Posts may not include email addresses, phone numbers, social links or links to unrelated websites; offers and updates are still allowed in Posts. Google's framing is that contributions should "reflect genuine experiences or define a business's permanent identity," not act as a marketplace.
TikTok opened a 400,000-app ad network with agentic lead-gen tools and a connector for Claude and Perplexity
Announced at Advertising Week New York on October 7, the TikTok Advertising Network lets advertisers buy across about 400,000 apps reaching 1 billion daily active users, including 44 of the top 100 US gaming apps. It shipped with agentic features: Buy Direct in-app checkout, a Shopping Assistant, agentic lead-generation tools, and a Model Context Protocol server that connects campaigns to Claude and Perplexity. TikTok says 84% of brands now advertise on the platform and 68% of its advertisers use its built-in AI tools.
Microsoft Advertising now syncs HubSpot lead stages into bidding and accepts LinkedIn lists of 10,000 companies
Microsoft's October update, published October 6, connects Microsoft Advertising directly to HubSpot so advertisers can send lifecycle stages such as marketing-qualified and sales-qualified leads back to optimize bids. Advertisers can also bulk upload up to 10,000 companies for LinkedIn profile targeting or exclusion (processing takes one to two days), and impression-based remarketing lists can now pool up to 20 campaigns, including connected TV as a source for Search. Microsoft published no performance data on lead quality.
Physical channel: USPS 2027 offers 40% off postage for businesses that have not mailed in two years, and the first sign-up windows open October 15
No new national consumer study on mail, print or out-of-home surfaced in the last seven days, so this week's physical-channel item is a deadline. In its 2027 promotions, filed with its July price change and still subject to Postal Regulatory Commission approval, USPS adds a "Direct Mail Discovery" promotion: a 40% discount on 5,000 pieces for eligible companies that have not mailed in the past two years, with up to 100,000 pieces covered from April 1 to September 30, 2027 and registration from February 15, 2027. The Tactile, Sensory and Interactive promotion (3% to 5%) opens registration October 15, 2026, and a 5% Integrated Technology promotion for NFC, AR, video in print and AI-driven pieces opens November 15.
Jaguar showed the car behind its rebrand: £130,000, and 78% of the 50,000 people who registered are new to Jaguar
Two years after the "Copy Nothing" rebrand, Jaguar unveiled the production Type 01 in New York on October 6, a four-door electric grand tourer from £130,000 (about $130,500 in the US). Jaguar's US average transaction price had been around $60,000 to $70,000. More than 50,000 people have registered interest and 78% of them are new to the brand; Jaguar's managing director described the old buyer as a shrinking, aging group of men in their fifties and the new target as tech professionals and design-literate urban buyers. Orders open in early 2027.
Starbucks' "third place" repositioning has delivered four straight quarters of growth
Adweek's October 6 profile of Brian Niccol's two years at Starbucks lays out the numbers behind "Back to Starbucks": global comparable sales up 7.9% in the fourth consecutive quarter of growth, stock up 30% since September 2024, more than $500 million spent on staffing, 1,000 stores refurbished, and 2025 advertising up 46% to $869 million. Operating margin fell from about 16% in 2023 to just above 11% because of that labor investment. Niccol describes the brand as somewhere to be with your community: "durable, sustaining, and highly human."
Ferrero is putting $50 million behind its 2027 Super Bowl after Kinder Bueno's debut lifted sales 140%
Marketing Dive reported on October 6 that Ferrero will spend more than $50 million across its portfolio for Super Bowl LXI, its largest single marketing investment, with Nutella appearing in the game for the first time. The bet follows Kinder Bueno's 2026 Super Bowl debut, which Ferrero says drove a 140% sales volume lift. Nutella also rode an unplanned moment: after the jar's appearance on the Artemis II mission, Amazon sales rose 20.2% and search volume rose 60.7% year over year.
Grocery Outlet reached 8.6 million shoppers in 12 days with an AI that chose what ran, plus an AI chat at the end of the ad
Grocery Outlet's Founders Day promotion ran on InMobi's Mobius system, which decides which ads to show across CTV and mobile and records why, reported October 5. Over a 12-day flight the campaign reached 8.6 million unique shoppers, with CTV video completion peaking at 98.7% against an 82.7% blended average. A wipe-to-reveal unit sent 6.6% of viewers to the store locator, and an AI companion at the end of the video, which talked about nearby stores and promotions, drew 35% engagement. All figures are vendor-reported, with no holdout or incrementality data.
The one-line read. The first week of October is about honesty and patience. Google now tells site owners in plain words that invented authors, AI headshots and false credentials are deception and a signal of a low-quality page, and its own engineers published how long things really take: about 20 hours to discover a new URL, about 30 days to recrawl a known one, and one to two days, sometimes months, for a title change to show. On the brand side, the strongest moves looked backward: the UAE's e& dropped its four-year-old name and went back to Etisalat for its 50th anniversary, and Nike raised brand spending 5% in a quarter when revenue fell 4%. The campaigns that worked were the ones that stayed on: Panda Express turned a Roblox stunt into a permanent destination, and Chili's grew its Triple Dipper from 7% to 15% of category sales through long-term creators rather than one-off posts. For a business that wants to be found, the condition is clear: real people, real names, and steady presence beat clever and fast.
[ AI search ]
Google now calls fake authors "a form of deception" in its public guidance
On October 1, Google added a new paragraph to its "Creating helpful, reliable, people-first content" page: avoid deceptive authorship information, and fabricating creator profiles "such as by using AI-generated headshots, made-up names, or false credentials" is a form of deception. Until now this lived only in the quality rater guidelines; it is now direction for site owners, and Google says deception is "a signal of a low-quality page" to its automated quality systems. The same revision defines main content and lists the four attributes raters weigh: effort, originality, talent or skill, and accuracy. This lands in the middle of the September spam update, which is still rolling out.
Google published how long crawling, indexing and title changes really take
At Search Central Live Deep Dive Europe, Gary Illyes shared Google's internal typical and slowest times for each step. Discovering a new URL takes about 20 hours typically and "weeks to never" at worst; refreshing a known URL takes about 30 days; a sitemap is processed in about 24 hours (up to 14 days). End-to-end indexing typically takes about 1.5 hours once crawled, but can take months or never. Snippet and title updates take one to two days typically and several weeks to months at the slow end. A Search Console removal takes about 2 hours.
Google is building ad units inside AI Mode, now testing sitelinks after text-link ads
Google is testing new sitelink formats inside AI Mode ads to see which earn the most clicks, spotted October 2. It follows a test last month of "Sponsored" ads written as plain anchor-text links inside the AI Mode answer, which look like organic answers. Search Engine Roundtable notes people rarely click links in AI responses, which is why Google keeps trying richer ad formats there.
Google Ads launched one-click AI localization of Search campaigns, including Latin American Spanish
On October 5, Google released a beta tool that duplicates an active or paused Search campaign into a new language. It translates headlines, descriptions, sitelinks, callouts and keywords, rebuilds image text overlays as translated assets, and can update landing page URLs. The original campaign stays untouched; the new one is fully independent, with budgets and bids copied exactly. Supported targets include Latin American and Castilian Spanish, Brazilian and European Portuguese, French (Canada and standard), German, Italian, Dutch and Polish.
From October 16, Google Ads lets some advertisers show a business name different from their domain
Google will allow a business name in ads that differs from the destination domain if three conditions hold: the name is the advertiser's recognized brand, there is a verified direct relationship between advertiser and domain owner, and the products are offered on that domain. Resellers, booking intermediaries, affiliates and secondary sellers are excluded.
Physical channel follow-up: "Final Notice" and "Act Now" get 45% of mail thrown away
No new national physical-channel study appeared in the last seven days. Lob's 2026 consumer survey (2,000 US adults, July fieldwork) is now being reported in full, and three findings were not in our first look. 45% of consumers discard mail that uses urgency phrases like "Final Notice" or "Act Now." Consumers under financial strain are 55% likely to buy from an unfamiliar brand, against 32% of financially secure consumers. And while 82% accept AI personalizing a mail piece, 67% oppose AI-generated images of people on it. Trust is up: 34% say mail feels more trustworthy than digital, 9 points higher than 2025, and 79% have taken a digital action after receiving mail.
On October 6, the UAE telecom group announced it is retiring the e& name at group level and returning to Etisalat to mark its 50th anniversary, saying the name "reflects our heritage, our strategy and our ambition for the AI era." e& had been adopted in 2022 to signal a pivot into a global technology investment group. The company is not retreating from growth: Q2 revenue rose 9% to $5.2 billion and subscribers grew about a third to 251.5 million. Some subsidiaries, such as e& International and e& Capital, still carry the old mark for now.
Nike raised brand spending 5% in a quarter when revenue fell 4%
Nike's quarter to August 31 showed revenue down 4% to $11.2 billion and China down 22%, yet demand creation spending rose 5% to $1.3 billion, driven by brand marketing. The increase sits inside a $2.5 billion, five-year cost-cutting program. CEO Elliott Hill wrote that the company is "redirecting investment toward the areas most critical to winning: product innovation, brand storytelling, consumer connection, sport and growth," and the money is going to sport moments and athlete-led stories rather than discounting.
Panda Express made its Roblox game permanent after a 5-point lift in purchase intent
Panda Express is relaunching The Panda Express Obby, an obstacle-course game on Roblox, as an always-on destination with monthly updates, seasonal courses for Halloween, fall and winter, and menu tie-ins. Its earlier gaming activations produced a 3-point lift in brand awareness, consideration up from 64% to 67%, recommendation up from 61% to 64%, and a 5-point rise in purchase intent over 30 days, with a 16% engagement rate on Roblox and 537,000 engagements on Fortnite.
Chili's doubled Triple Dipper to 15% of sales with long-term creators, not one-off posts
At Advertising Week New York on October 6, Chili's said Triple Dipper grew from 7% to 15% of its category sales through ongoing influencer partnerships, including mukbang content with creators such as Keila Pacheco, and that a low-production Lizzo video beat the high-budget original spot. Disney said creator content is now more than 40% of its campaign assets, up from under 10% five years ago. Alaska Airlines moved to long-term creator contracts for budget efficiency and found creators prefer them. All three now bring creators into planning before a brief exists.
Ecover made four T-shirts you are not supposed to wash
For Sustainable Fashion Week (ending October 4), Ecover and Uncommon Creative Studio made The Unwashables: four "lucky" T-shirts dyed with 52 crushed four-leaf clovers, rusted horseshoes, wishing-well water and dust from a 6,000-year-old meteorite. It is built on Ecover's research that 30% of Brits avoid washing lucky clothes and 54% believe what they wear affects their luck. The shirts are not sold; a few are given away through social contests, all pointing to one message: make the clothes you love last longer.
The one-line read. The week opened with the clearest picture yet of what AI search is worth and who it rewards. IDHL's 13-month study of 1.5 million AI sessions found AI search traffic up 392% and transactions up 553%, with ChatGPT carrying 91% of it and Reddit now the most-cited third-party source. The same week a federal judge threw out the publishers' case against Google AI Overviews with the line "an expectation is not an agreement," and Google told a room of SEOs in Europe that it filters 40 billion spam pages a day and now uses AI to find AI-made spam. On the paid side, OpenAI added a visual ad format to ChatGPT and published its first measured results, including a 15.3% lower cost per acquisition for WeightWatchers, while Apple's iOS 27 quietly stopped The Trade Desk from serving ads in Safari. The condition for any business that wants to be found: the AI channel now converts, nobody is owed a click, and the only durable position is being the source the model cites.
[ AI search ]
AI search traffic is up 392% in 13 months and transactions are up 553%, with ChatGPT carrying 91% of it
IDHL's State of AI Search report, published September 30, tracked 1.5 million AI search sessions across 175 websites and 20 industries over 13 months. AI search traffic grew 392% year on year, transactions grew 553% (21,141 in total), and revenue grew 312%, which IDHL calls roughly 14 times the growth rate of organic search (28%). ChatGPT drove 91% of AI search traffic. Reddit is now the single most-cited third-party source across industries, after Reddit results were chosen 23 billion times inside Google searches last year. Home and Garden was one of the fastest-growing sectors at 588%.
A federal judge dismissed the Penske and Chegg suits over AI Overviews: "an expectation is not an agreement"
On October 1, Judge Amit Mehta dismissed Penske Media's antitrust claims against Google over AI Overviews, and a parallel case from Chegg. Penske argued that publishers supplied content in exchange for referral traffic and that AI Overviews broke that bargain; its complaint cited a 34.5% click-through reduction from AI Overviews and Google's 89.2% search share. The court found publishers "failed to plead any actual agreement" promising any amount of traffic. Mehta acknowledged harm to "journalists, educators, and other online creators" but declined to provide a remedy. Publishers are now turning to legislation, European regulators and direct licensing.
Google filters 40 billion spam pages a day and now uses AI to find AI spam
At Search Central Live Deep Dive in Europe on October 5, Gary Illyes said Google filters out 40 billion spam pages a day and that it is "using AI to help catch more." Google gave three reasons quality updates are now more frequent: more content formats, enough volume that evaluation has moved to the page level rather than the domain level, and a surge in cloaking, doorway pages, scraped content and scaled low-effort content. Scaled content is now a greater risk than link spam. The session followed the October 1 update to Google's helpful content document, which added a section stating that main content quality is the most critical factor raters assess. This is a follow-up to last week's "manually fact-check AI content" guidance.
The "AI Overview tracking parameter" was a Chrome extension, not Google
On October 2, SEOs reported links inside AI Overviews carrying a new parameter that would have let site owners tell AI Overview clicks apart from regular search clicks. Tom Pool, who spotted it, later confirmed it came from a Chrome browser extension rewriting URLs, apparently to resell the data, and not from Google. Google still gives no way to separate AI Overview and AI Mode clicks in Search Console.
Google Local Services Ads now hide the phone number behind a hover, and missed calls became billable October 1
Search Engine Roundtable reported on October 2 that Local Services Ads now show a "Get phone number" button instead of the number, revealing it only on hover. The change arrived alongside an October 1 policy that makes certain missed calls billable, and a separate bug where ads served outside their targeted area. Earlier reports also found LSAs showing the Google Business Profile number rather than the tracking number set in Google Ads.
OpenAI added visual ads to ChatGPT and published the first measured results: 15.3% lower cost per acquisition for WeightWatchers
On October 5, OpenAI introduced a visual ad format that will be tested in the US later this month during image generation, with labels and separation from the user's own image. It said ChatGPT reaches 1.2 billion people a week, and named its measurement stack: conversion tracking through Hightouch, Tealium and LiveRamp, and attribution through AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava and others. DoubleVerify and Integral Ad Science are running brand suitability pilots that evaluate conversational context without reading private chats. Early results: WeightWatchers saw a cost per acquisition 15.3% below its paid-search benchmark (DV Rockerbox), Dose found 67% of incremental purchases came from net-new customers, and 93% of Portland Leather's ChatGPT visitors were new. Separately, Similarweb reports that ads shown in the first four turns of a conversation earn a 48% higher click-through rate, against an overall ChatGPT ad click-through rate of 0.68%.
Google Ads will start pulling coupons off your website and into your ads on October 12, unless you opt out
Starting October 12, Google Ads will automatically detect promotions on advertisers' websites and attach them as promotion assets to Search and Performance Max campaigns. It applies to campaigns with linked location assets and no existing promotion assets, re-verifies the offer within 24 hours, and is on by default; the switch is in account-level automated asset settings. The same week Google said it will let eligible advertisers use recognized brand names that do not match their destination domain.
Apple's iOS 27 blocks The Trade Desk from serving ads in Safari, and three weeks later there is no fix
Apple added adsrvr.org, The Trade Desk's core ad request and delivery domain, to Safari's unconditional block list in iOS 27, released September 14, alongside identity domains from UID2, ID5, LiveRamp, Audigent and Permutive. The Trade Desk filed an urgent request on September 21; on September 28 WebKit's John Wilander would only say he would notify them "if and when" a change was available. AdExchanger reported on September 29 that the block was still in place, and as of this week it remains unresolved. The Trade Desk says the domain handles ad delivery, not identity.
Physical channel: 68% of Americans notice billboards, 59% notice signs, and 35% have researched a product on their phone after seeing an out-of-home ad
YouGov's US out-of-home effectiveness analysis, drawn from YouGov Profiles data collected July 2025 to July 2026 and charted by MarketingProfs, finds 26% of Americans notice out-of-home advertising several times a week and 21% multiple times a day; only 3% say they never notice it. Billboards are the most-noticed format (68%), ahead of signs (59%), point-of-sale displays (31%) and posters (31%). After seeing an out-of-home ad, 35% have researched the product or service on their phone, 20% have discussed the brand and 19% have made a purchase on their phone.
Mark Ritson's seven questions for Jaguar: positioning has to be stated in the customer's words, not the agency's
With Jaguar's £120,000 Type 01 finally unveiled, Ritson's October 5 column lists the seven questions the launch must answer. Among them: who is the target, given that "young, design-minded, cash-rich" is not a real segment; what share of orders are conquest versus retained owners, since Jaguar projected keeping only 10% to 15% of them; how many of its distinctive assets (the leaper, the growler, British Racing Green, the long bonnet) survive; and whether Jaguar can state its positioning in customer language rather than words like "exuberant" and "fearlessly creative." His core argument: Jaguar confused a change of powertrain with a change of identity, where Porsche carried its heritage into the Taycan.
LinkedIn: 85% of B2B marketers believe Gen Z buyers favor brands validated by their professional contacts
LinkedIn research reported by Marketing Week on October 5 found 88% of UK B2B marketers say future growth depends on influencing younger decision-makers, 85% believe Gen Z buyers favor brands validated by professional contacts, and 43% name expert-created content as the most effective way to reach them. In the same roundup, YouGov found British women on Reddit nearly doubled from 11% to 19% in four years, with 40% of them Gen Z and 45% millennials, and Gartner reported 18% of marketing leaders have already cut junior roles to automation.
KFC turned three years of "Believe in Chicken" into a 30-minute game show with a £50,000 prize
"Don't Crack," by Mother and directed by Pablo Rochat, premiered on KFC's UK YouTube channel on September 30. A March callout ad drew 9.8 million QR scans and 50,000 applications; 30 contestants aged 22 to 65 (a police sergeant, a tube driver, a GP, a self-described witch) face four escalating tests of devotion, from a blindfolded hunt for golden eggs to shedding a single tear in a giant nest. One walks away with £50,000 and the title of Ultimate Believer. It is the closing chapter of the platform KFC launched in 2024.
Reese's is searching 150 streets for America's "Trick-or-Treatiest," with Zillow throwing the block party
Announced October 1, Reese's narrowed the country to 150 streets using candy and costume purchases, decoration data, walkability, Census data, community participation and a survey of 15,000 residents. A first-ever Reese's Halloween Council (Christina Ricci, Kathy Najimy and creators Rachel Iwanyszyn, Zaire Alford and Macy Blackwell) will name the winner on October 21. The prize is a Halloween block party on the winning street, hosted with Zillow, with Reese's treats and a trick-or-treating experience.
Google AI Overviews now show on 80% of branded queries, up from 26% in September
DemandSphere data reported on September 30 shows AI Overviews on branded searches roughly tripled during September, from 26% to 80%, with the surge accelerating late in the month. Separately, Chris Long tested about 100 brands and found AI Overviews on 93 of them, including Reddit, Salesforce, Amazon, Adobe, Samsung, Apple and Poland Spring. Most sat mid-page or lower rather than at the top, with Adobe's appearing above its own company listing. Google's own name and news publishers were the exceptions.
Links inside AI Overviews are starting to open AI Mode instead of websites
Google has begun testing underlined text in AI Overviews that launches an AI Mode conversation rather than sending the user to an outside page, shown on video by Gagan Ghotra. At the same time, Peec AI's Malte Landwehr reported the share of sampled overviews containing embedded external links was near zero through September 10, then jumped to 26.2% by September 20. Search Console does not separate clicks from these AI features, so publishers cannot tell whether the net effect is more or less traffic.
Google now calls manual fact-checking of AI content "critical"
Google updated its guidance on using generative AI content on October 1, its first revision since October 2025. It now says "It is critical to manually factcheck and review all AI-generated content" before publishing, explaining that generative models predict likely words rather than retrieve facts. The guidance is not a ban: AI-assisted pages are fine when useful, original and accurate, but mass-producing pages without added value can still violate the scaled content abuse policy. Observers tied the unusually strong wording to the September 2026 spam update.
ChatGPT Ads added bulk product feeds and secondary conversion tracking, and now runs in 60+ countries
OpenAI's September 30 update lets advertisers build multiple product feed campaigns and ad groups from a spreadsheet, track product review status in a Products tab, and report conversion events beyond a campaign's main goal through the public Insights API without changing optimization. Account, brand and legal names are now managed separately. The platform launched across Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan and now runs in more than 60 countries.
Yext is buying Flamel.ai to run local ads on Google, Meta and ChatGPT for multi-location brands
On September 30 Yext announced a "multiplayer agent harness" in its Scout product, a shared workspace where corporate teams, franchisees and local managers direct the same AI agents across search, AI search, listings, reviews and social. The same day it agreed to acquire Flamel.ai, which runs localized paid campaigns on Google, Meta and ChatGPT for brands with hundreds or thousands of locations. Once closed, Scout will flag locations losing to local competitors and the agents will shift spend there. No price was disclosed; the deal is expected to close by January 31, 2027.
Contentstack released a free AI visibility report covering ChatGPT and Gemini
Contentstack Canoe, launched October 1, scans a site, identifies the business, products and up to ten competitors, generates buyer questions and reports how ChatGPT and Gemini answer them, in about two minutes. The free plan runs monthly with a crawler-readability audit and ranked fixes. The $279 per month Growth plan adds daily or weekly runs, editable questions, Perplexity and Claude, and an MCP server so an AI agent can draft fixes and rerun the check.
Kargo opened agentic media buying to small teams, with a 21% ad recall lift on an Ad Council test
Kargo made its self-serve platform Karlo available on October 1. Advertisers start from an existing marketing document or a plain-language prompt to set goals, build audiences and launch campaigns across connected TV, social and the open web from one dashboard, without a dedicated analyst. On an Ad Council campaign for "Love, Your Mind," Karlo delivered a 21% lift in ad recall and a 10% lift in aided awareness.
Marketing Week: Boots is on its third positioning in five years, and that is the problem
Helen Edwards argued in Marketing Week on October 2 that Boots, a 180-year-old brand, has changed its positioning, line and campaign three times in five years: a "feel good" wellbeing idea, "With you" in 2021, and now "Give it some Boots." She sets it against Nike ("Just do it" since 1988), Red Bull ("Gives you wings" since 1987) and L'Oréal ("You're worth it" since 1971), and argues marketers undervalue consistency. This follows up our September 10 item on the Boots relaunch.
IGT unveiled a new identity at G2E in Las Vegas after its merger with Everi
IGT revealed a bolder, modern brand identity at the Global Gaming Expo in Las Vegas, which ran September 28 to October 1. The company describes the new look as "fluid, frictionless and magnetic" and anchors it on four pillars: integrated casino floor solutions, high-demand gaming content, operator expertise and customer success. CEO Hector Fernandez said the identity "signals who we are becoming." The rebrand marks IGT's next chapter after combining with Everi Holdings.
Segafredo launched "Take Your Shot," turning the espresso into a moment before action
Segafredo Zanetti launched its first global platform from BBH London on October 1, starting in Italy and expanding across Europe on TV and social, with out-of-home later in 2026. The idea is the short pause before you act: the espresso as a trigger for people who "decide not to wait for opportunities," in CEO Pierluigi Tosato's words. It comes with a simplified logo, revived color palette and new typography, plus a launch film directed by Rodrigo Inada through Iconoclast.
Sharps sold fitted wardrobes with two sisters who refuse to be separated
UK fitted furniture brand Sharps launched a new platform, "It's What's Inside That Counts," with agency Neverland and media agency Medialab. The launch film "Sisters," directed by Molly Manners in 60 and 30 second cuts, follows two young siblings who sneak between neighboring bedrooms; the new wardrobes are built on both sides of the wall in a way that keeps them connected. The rollout includes DRTV, social, a new experiential Design Studio retail concept and an updated identity, aimed at standing out in a commoditized category.
Google Maps now requires a sign-in to read every review on a Business Profile
As of September 30, Google Maps asks users to sign in before they can see all reviews on a listing, sort them, leave one or view all photos. The prompt reads "Sign-in to get the best of Google Maps." Search Engine Roundtable's Barry Schwartz reads it as a move against "bots, scrapers, AI platforms and other third-parties" pulling review data. Several local SEOs spotted it on X and LinkedIn the same day. Sources: Search Engine Roundtable.
Google's AI payments to publishers come to about 0.1% of their ad revenue
The Information reports that Google's AI contribution pilot now pays about 100 publishers for content used in AI Overviews, AI Mode and Gemini. Small sites have received under $1,000 over several months, mid-sized publishers $50,000 to $60,000, and large ones $1 million or more a year. For several small and mid-sized sites, the payments come to less than a tenth of 1% of their advertising revenue, and publishers say the monthly amounts change without explanation. Sources: Search Engine Roundtable, October webmaster report.
WFA: 73% of major brands watch their AI visibility, but only 8% have a clear plan
The World Federation of Advertisers surveyed 54 senior marketers at 46 brand owners with $93 billion in combined ad spend, released September 30. 96% now use generative or agentic AI, up from 45% in 2023. 85% are moderately or very concerned about brand visibility and control in AI answers and LLM advertising, and 73% monitor AI visibility, but only 8% have a clear strategy for it. Sources: WFA.
Gen Z now trusts ChatGPT over Google for product research, 31% to 12%
A Talker Research survey of 2,000 US adults for Spreedly, fielded September 3 to 8 and released September 30, found 31% of Gen Z name ChatGPT as their most trusted product-research source against 12% for Google Search. Across all ages Google still leads at 25%. 63% would let an AI assistant choose brands for them if it reliably saved 20%, but only 20% would let it spend without approval, and 64% refuse outright. Sources: Finopotamus.
Amazon folded its ad tools into one "Amazon Ads Agent" with three campaign types
At unBoxed on September 29, Amazon rebranded its Ads Console and Amazon DSP as Amazon Ads Agent, with a chat interface and three campaign types: Sponsored Ads, Display, Video and Audio (DVA+), and Full-Funnel Campaigns. Amazon says DVA+ lets an advertiser pick a goal, add creative and a budget, and launch within minutes while AI handles optimization. The minimum daily budget is $100, US only for now. Sources: MediaPost.
Meta's Muse agent hit 5 million US downloads in 22 days and launched a small-business version
Follow-up to the September 29 item. Meta Muse reached 5 million US downloads 22 days after its September 8 launch, faster than ChatGPT (56 days), Grok (103) and Claude (492). Meta says more than 3 million people submit prompts weekly. On September 29 Meta announced Muse for Small Business, connecting to Shopify, Slack, Stripe, Canva, Dropbox, Notion and QuickBooks; it asks for approval before purchases or sending email. Sources: Storyboard18.
Kroger forecast e-commerce conversion from creative alone with 81% accuracy
Kroger, Vidmob and MMA Global scored more than 1,900 video and image ads from Meta and DV360 campaigns on messaging, story structure, branding and on-screen people. Reported September 30: the scores predicted e-commerce conversion with 81% accuracy, creative that followed the recommendations converted about 4x better on average, and moving spend to higher-scoring creative could produce more than 2x conversions from the same budget. Sources: Marketing Dive.
Holiday email volume is costing brands trust, especially with Gen Z
Validity's Holiday 2026 Consumer Survey, released September 30, found nearly half of Gen Z say holiday email practices reduced their trust in a brand or their intent to buy, and that volume, not content, is the main cause. 36% of consumers have missed important non-promotional messages because of the flood, 47% say trust decides whether they open a holiday email, and 22% would use a temporary "opt-down" or snooze. Sources: Yahoo Finance.
McAfee repositioned from "antivirus" to "Safer Than Safe"
McAfee launched a corporate rebrand on September 29 with the line "Safer Than Safe" and an R/GA campaign led by a 60-second spot. The aim is to move the brand from what people remember (antivirus software) to what it now sells (broader digital safety against scams and AI-driven threats). The product side changed too: clearer alerts that tell users what specific issue was found. Sources: Wisevoter.
Wayfair's "Wayfair Delivers" put a real NFL family in its living room and grew sports spend fivefold
Wayfair launched "Wayfair Delivers," a values-led platform built on quality and reliable service, with a 30-second spot that debuted on Monday Night Football. It stars quarterback Matthew Stafford, his wife Kelly and their four daughters, contrasting his football skill with chaos at home. The spot was made in-house with Paramount Media Labs, and Wayfair raised spend on football networks to five times prior levels. The platform promotes two proof points by name: Wayfair Verified and Wayfair Rewards. Sources: Marketing Dive.
Standard Chartered's "Move Ahead" used AI production to build scenes it could not film during the F1 season
Standard Chartered launched its global "Move Ahead" campaign on September 28, built on its Formula 1 partnership and the idea that success is prepared long before the moment arrives. The bank says the work blended human creative direction with AI-enabled production to create racing imagery that would have been hard to shoot mid-season, and it runs across digital, social and owned channels. Sources: Standard Chartered.
The one-line read. Google's AI search has stopped being a side feature. TIME reports AI Mode now has more than a billion monthly users and AI Overviews appear on 40% of US searches, and people in AI Mode type queries three times longer and rarely check the answers. On September 28 Google gave everyone the ability to tell AI Mode to watch for something and report back. New research on 34,960 AI answers found that a brand whose domain is never retrieved is named in fewer than 4% of answers, compared with 49% to 58% once the domain is cited. Consumers are pushing back at the same time: 43% of Americans in a new Alchemer survey say they would pay more for guaranteed access to a human, and a global Usercentrics study found that consumers who hand data to AI assistants reluctantly outnumber willing ones more than two to one. The strongest brand work this week was about the customer, not the company: Thorne's consumer-first platform came before a $3.8 billion sale, and e.l.f. keeps paying for music that promotes emerging artists. For a business that wants to be found in search, the rule is to get the domain cited first, write for the long question, and state plainly that a person is on the other end.
[ AI search ]
Google's AI Mode passed a billion monthly users, and people in it write queries three times longer and rarely check the answers
TIME's September 29 report says AI Overviews now appear in 40% of US searches, AI Mode has more than 1 billion monthly users, and 60% of Americans read AI summaries in search results, compared with 49% who use chatbots. Google has moved AI features out from behind "Show More" buttons and this month began placing AI Mode responses inside AI Overviews. Researchers describe AI Mode users as showing a "high level of trust" and doing little to verify answers. (TIME, Search Engine Roundtable: AI Mode inside AI Overviews)
AI Mode can now watch the web for you: monitoring rolled out to all users globally on September 28
Google moved AI Mode monitoring from Ultra and Pro subscribers to all users worldwide through the Google app. Users describe what to watch, and Search "will continuously check across changing info" across websites, forums, social posts and the 60 billion-product Shopping Graph, then sends a notification when conditions are met. Google's examples include new restaurants opening nearby and price drops. (Search Engine Roundtable)
Retrieval comes before recommendation: brands whose domain is never retrieved are named in fewer than 4% of AI answers
A September 19 paper by Benjamin Tannenbaum analyzed 34,960 unbranded prompt observations across 75 projects and 2,854 prompts in GPT and Gemini from June to September 2026. With no domain presence, brands were mentioned 2.8% of the time in GPT and 3.8% in Gemini. When the domain was cited, mentions rose to 49.0% and 58.4%. When it was cited and the engine also ran a branded follow-up search, mentions reached 91.4% and 100%. Prior visibility carried over: a brand mentioned before and retrieved again was named 80.5% and 83.7% of the time. (arXiv)
Google is again testing AI Overview citations at the bottom instead of the right side
Glenn Gabe spotted a desktop test that removes the right-side citation bar and puts larger citation cards below the answer, behind "show more," with inline hover links kept. It is Google's third test of this layout, after September 2025 and March 2026. Gabe said the bottom links "look much more enticing down there." Hiding them behind a click may still reduce how often they are seen. (Search Engine Roundtable)
Google Business Profile added a policy against misusing the support portal, with a lasting penalty
Google's GBP policies now ban false or misleading reports and bad-faith complaints through the support portal, and warn that "repeated unfounded submissions can result in rejection of all further complaints without assessment." Search Engine Roundtable also noted a test of two rows of action buttons in mobile local panels and a Maps bug placing service-area business pins in the ocean. (Search Engine Roundtable: policy, local panel test)
Braze launched self-service AI decisioning agents and an agent that checks every campaign against brand rules before it sends
At its Forge conference on September 29, Braze announced Decisioning Studio Go, which picks each recipient's message variant, send time, day and frequency within marketer-set guardrails. Braze says one agent can test more than 1 billion combinations at once, and general availability is October 14. Agentic Standards checks campaigns for setup, copy, links, personalization, compliance and brand guidelines, marks each pass, warning or fail, and applies fixes only with human sign-off. Operator Connect lets marketers work in Braze from Claude, ChatGPT, Copilot and Snowflake Cortex. Braze award winners reported a 28% lift against a holdout (Nestlé Purina) and a 9.07% conversion lift (Compare the Market). (Braze investor release, MarTech, Agile Brand Guide roundup)
"Resigned consent": for every consumer who willingly lets an AI assistant use their data, more than two let it in reluctantly
Usercentrics' State of Digital Trust Report 2026 (11,000 consumers in seven markets including the US, run by Sapio Research) found only 7% are fully comfortable giving an AI assistant access to their data with no conditions, while 17% grant access despite discomfort. Nearly six in ten are uncomfortable overall. Resigned consent was 10% in the US and 24% in Germany, and financial accounts drew 64% discomfort. Usercentrics argues that a reluctant yes and a real yes look identical in a consent rate but behave differently later. (Usercentrics, City AM)
43% of Americans would pay more for a product that guarantees access to a human, and daily AI users feel it most
Alchemer's 2026 AI & Customer Expectations Report, a September 29 survey of 2,009 US consumers, found 43% would pay more for guaranteed human support, 44% say AI has made it harder to reach a person, and 37% have received AI answers that misunderstood their question. Daily AI users were more than 2.5 times as likely as non-users to say they would definitely pay extra for human support, and 40% of them call AI customer experience impersonal, compared with 22% of non-users. (MarTech Series, FinancialContent / Business Wire)
Thorne reached a $3.8 billion sale after moving its message from "Thorne-first" to "consumer-first"
Supplement brand Thorne, bought by P&G for $3.8 billion in August, built its brand platform with Redscout using digital diaries, a nationally representative quantitative study and small-group interviews. The finding: consumers felt overwhelmed by supplements, and scientific claims mattered less than feeling confident in their choices. The platform line is "Thorne rewards curiosity with excellence," the internal brief is "Know me, show me," and the "Now I Know" campaign paired named experts with specific life moments (Misty Copeland on perimenopause, Cade Cunningham on sports performance). Chief Growth Officer Mary Beech described the shift as "consumer-first versus being Thorne-first." (Marketing Dive)
CMO tenure has fallen 35% since 2010, and fractional marketing leaders tripled in four years
The Findem and CMO Huddles study of more than 13,000 US CMOs found median tenure fell from 4.0 years for the 2010 cohort to 2.6 years for the 2022 cohort, compared with 51 months for CEOs. Venture-backed companies average 2.6 years and public companies 4.0. Fractional or interim CMO appointments rose from 2.9% in 2020 to 8.8% in 2024 and held near 8% since. Drew Neisser called the role "the pressure cooker of the C-suite." (Agile Brand Guide, MediaPost)
e.l.f. released a second album of emerging artists, after the first drew more than 500,000 streams and 10 million views
e.l.f.'s entertainment arm, e.l.f. Made, released "Mirror Mix," seven tracks from emerging artists across pop, R&B, Latin and rock, distributed on Spotify, Apple Music and inside e.l.f.'s Roblox experience. Genius promoted it through its Open Mic series with a live performance on September 30. The 2024 album drew more than 500,000 Spotify streams and 10 million views. The reasoning: "one in three independent artists are challenged by lack of exposure," so the brand gives exposure and gets content people choose to consume. (Marketing Dive)
Twisted Tea put Realtree camo on its cans and merchandise while holding more than 85% of its category and facing headwinds
Twisted Tea's fall campaign with Realtree puts camouflage on cans and cases across the line, adds a co-branded merchandise store (hoodies, vests, chairs, duck calls) and a "Tea Drop" ad spanning TV, streaming, social, on-premise and retail displays. The brand holds more than 85% of malt-based hard tea while the category struggles, and Boston Beer's new CMO starts October 12. Senior brand director Erica Taylor: "Our fans are the kind of people who would rather be outside doing something than sitting around talking about it, and Realtree is already part of that world." (Marketing Dive)
The one-line read. The AI audience is spreading out while the AI clicks are piling up in one place. Comscore showed last week that ChatGPT now handles only half of AI prompts, yet BrightEdge's new data says ChatGPT sent 95.1% of all AI referral traffic to websites in August and more than doubled that traffic since January, with Gemini, Perplexity and Claude sharing the remaining 5%. Google is testing AI Overviews that expand on their own, Bing is trying on Google's "AI Mode" name, and Shopify just let browser agents complete checkout on a million stores as Meta's Muse agent passed 2.5 million downloads in two weeks. The measurement stack is racing to catch up: ChatGPT Ads attribution, out-of-home exposure inside LiveRamp and TransUnion, and a single AI Max report that finally links search term, ad and landing page. On the brand side, this week's winners picked a seat and sat in it, Subway in the middle of fast food and Target on design, while Dave & Buster's showed what a year without a marketing leader costs in comparable sales. For a business showing up in search, the condition is: be citable in every assistant, be readable by an agent, and give your marketing a fixed address people can find again.
[ AI search ]
BrightEdge: ChatGPT referral traffic to websites more than doubled since January and reached a 95.1% share of AI referrals in August
BrightEdge's analysis of AI referral traffic from January through August 2026 shows ChatGPT referrals up 101% over the period and up 36% in August alone. ChatGPT's share of all AI-assistant referrals hit 95.1% in August, its 2026 high, after dipping to 89.5% in April. Gemini held 2.4%, Perplexity 1.1% and Claude 0.8%, and in August Claude's referrals fell 15% and Perplexity's fell 17% month over month. Total AI referral traffic grew about 73% in under a year. BrightEdge CEO Jim Yu: ChatGPT "is also sending substantially more users to the web than it was at the beginning of the year." (BrightEdge via GlobeNewswire)
Bing is testing Google's "AI Mode" name on its own search tabs, and Google's AI Overviews are being tested with automatic expansion
Microsoft is testing an "AI MODE" tab in Bing alongside All, Images, Videos and Maps, and an AI Mode entry point in Edge; both send people to bing.com/copilotsearch, the same destination as Copilot Search. It is the second rename test this year, after Bing tried "AI Overviews" on its Copilot answer box in August. Separately, Google's "Loading..." button test in place of "Show more" under AI Overviews, noted yesterday, was documented again on September 28 by Sachin Patel and Khushal Bherwani, with the AI Overview expanding into a longer, AI Mode style answer without a click. (Search Engine Roundtable on Bing, Search Engine Watch, Search Engine Roundtable on AI Overviews)
Google Ads will show one AI Max report linking the search term, the creative shown and the landing page reached, and AI Brief now takes instructions in seven more languages
Google announced a unified AI Max for Search report that, for each ad interaction, shows which search term triggered the ad, which assets the person saw and which page on the advertiser's site they landed on. There is no launch date yet; Google says details and availability come later in 2026. AI Brief, the Gemini-powered field where advertisers steer messaging, creative guidelines and audience before a campaign runs, expanded to Dutch, French, German, Italian, Japanese, Portuguese and Spanish. The timing lands as Google finishes auto-migrating campaigns with campaign-level broad match or automatically created assets into AI Max by the end of September, with no rollback. (Search Engine Land, Search Engine Roundtable, MediaPost)
Shopify let browser agents complete checkout on its merchants' websites, days after Meta's Muse agent passed 2.5 million downloads
On September 28 Shopify said browser agents can now read, update and submit checkout on the websites of all eligible merchants, with the buyer's authorization, through WebMCP support for checkout including Shop Pay. Named agents include Meta's Muse and Instinct's personal agent. It follows Shopify's September 21 move to switch on agent checkout by default for roughly one million US merchants (opt out in Shopify Admin under Sales channels, Agentic). PYMNTS cites survey data that 56% of consumers would let an agent search and compare products and 35% would give one access to saved payment credentials. Muse, launched September 8, hit 730,000 downloads in five days, overtook ChatGPT as the top free iOS app in the US on September 18, and reached 2.5 million downloads by September 21 per Sensor Tower, against 400,000 for Claude and 200,000 for Grok in the same window. (PYMNTS, Yahoo Finance, CNBC)
AI-channel attribution arrived from two directions: Branch now attributes ChatGPT Ads across app, web and desktop, and Fairing asks buyers which podcast, creator or AI assistant sent them
Branch announced on September 24 that, as a ChatGPT Ads measurement partner, it can connect activity from ChatGPT campaigns to installs, app opens, conversions and downstream outcomes on mobile apps, websites and desktop. Fairing's Advanced Attribution, in beta since September 10 with general availability planned for January 2027, adds a dynamic follow-up to the "How did you hear about us?" survey: pick Podcast, YouTube or AI and a second question asks for the specific show, creator or assistant, with managed auto-suggest lists that turn free text into structured labels. Fairing says early users got up to 50% more usable attribution signal from the same order volume, a figure it has not independently documented. (Branch via PR Newswire, PPC Land on Fairing)
Clear Channel Outdoor put billboard exposure into LiveRamp and TransUnion, so out-of-home can be deduplicated against CTV, social and audio in one model
On September 24 Clear Channel Outdoor expanded its RADAR data platform so out-of-home audience, exposure and delivery data flows into LiveRamp's Cross-Media Intelligence, where OOH exposure is measured and deduplicated against other media, and into TransUnion's marketing mix modeling and multi-touch attribution products, where RADAR exposure is resolved to customer-level identity. Clear Channel is the first OOH company inside LiveRamp's tool. No match rates or validation results were published. CMO Dan Levi: marketers "need to be able to see its contribution within the same measurement frameworks they use across the rest of their media plans." (Clear Channel Outdoor, PPC Land)
Yard signs and mailers land hardest with Gen Z: 23% say physical materials carry the most weight, against 8% of Gen X and no Boomers
A UPrinting survey of 1,000 US adults, run on Pollfish in August 2026 about political campaign materials, found 23% of Gen Z saying physical materials (signs, mailers, bumper stickers) carry the most influence on their view of a candidate, against 15% of Millennials, 8% of Gen X and 0% of Boomers. 21% of Gen Z read a street dominated by one candidate's yard signs as proof of real momentum, versus 5% of Boomers. 43% of Boomers said the materials had no effect on them at all, roughly double Gen Z's rate. Caveats: the sample was unweighted, skewed young and 61% female, and respondents were not asked their politics. (Stacker via KESQ)
Subway's US CMO says the growth is "in the middle": between value fast food and premium healthy, not against either
Jeff Klein, who joined Subway in March and was elevated to US CMO, told Marketing Dive on September 28 that the brand's room to grow is the gap between traditional quick service and premium healthy chains: "There's a lot of space in the middle. It's a big step from Subway to some of the uber-healthy brands when it comes to price." The moves: a "Feed Your SUBconscious" brand platform with Will Arnett, creative and media refocused on core guests, fewer scattered movie tie-ins, and a second Arnett batch in spring. The context is why it matters: 2025 revenue fell more than 6%, 729 restaurants closed in 2025 and more than 3,000 since 2021, leaving about 19,000. Klein: "I don't think we're going to fully reverse the challenges with incrementalism." (Marketing Dive)
Dave & Buster's went more than a year without a CMO, and the new one says the result was off-target messaging, a shifting promo calendar and unreliable measurement
Dave & Buster's named Gut as creative agency of record and Mediaplus as media agency on September 28, replacing Deloitte Digital and Dentsu, with both starting October 1. The review began in spring under new CMO Jeremy Tucker, who arrived from AutoNation in June after the 251-location chain had gone more than a year without a marketing chief and cycled through three CEOs since 2021. Revenue fell 2.4% to $544.1 million and comparable sales fell 2.9% in the quarter ended August 4. The stated fix is nationally scaled campaigns tied to local market relevance and a data-driven media plan. (Marketing Dive)
Target's "Inside Out" makes the shopping bag invisible and the customer the design, as Q2 comparable sales rose 3.8% and traffic 3.6%
Target's fall campaign, launched September 23 and running through October 31 across broadcast, streaming, radio, digital and social, shows real customers pulling Target purchases out of invisible shopping bags that materialize on screen, set to Raye's "Joy." The point is to move Target's design story from the shelf to the person: VP of Creative Scott Swartz says the brand is "putting guests at the center." It lands mid-turnaround, with Q2 2026 comparable sales up 3.8%, traffic up 3.6%, a $1 billion operational investment and a new chief marketing and customer officer, ex-Hilton CMO Mark Weinstein. (Marketing Dive, Retail Dive)
Brooklinen's first celebrity campaign was made in-house, with four unrelated famous people and a media plan timed to the SNL premiere and NBA tip-off
Brooklinen's "Best. Sheets. Ever." campaign, launched August 27 and profiled September 25, is the brand's first celebrity effort and its largest omnichannel push, built in-house with director Barbara Anastacio rather than through an agency. Comedian Robby Hoffman, SNL's Ben Marshall, Alana Haim and Knicks forward Mikal Bridges each get a bed-bound scenario (a therapy session, dodging plans, waking up, a work call) in a 30-second hero spot with 15-second cutdowns and behind-the-scenes content of the talent folding fitted sheets. Media runs on linear TV, connected TV, YouTube and Meta through fall, timed to the September 26 SNL season premiere and the October 20 NBA opener. CMO Laura Jacobs: "We talk about the bedroom as a place that's full of life where lots of things happen." (Marketing Dive)
Jeep's "This Is Jeep Country" ties the brand to America's 250th and to 85 years of heritage, with US sales up 6% in a market that fell 0.3%
Jeep's new platform from agency Highdive, profiled September 25, runs across NFL media, digital and connected TV on Disney+, Paramount+, Max and Peacock, with all seven models in the portfolio and dedicated African-American and Hispanic spots. The tone is deliberately bolder and more patriotic, tied to the country's 250th anniversary and Jeep's wartime origin; Highdive's Jeremy Adams: "America was the Jeep brand's first customer." It rides real momentum: Jeep's US sales rose 6% in Q2 2026 while the industry fell 0.3%, inside a roughly $70 billion five-year Stellantis plan. Jeep's advertising head Lindsay Dyer Fifelski: "Jeep Country lives where our owners live. It's an attitude." (Marketing Dive)
The one-line read. The weekend made Google's September spam update real, with big drops across verticals and AI-generated content sites among the losers, and it lands on a search channel that is already shrinking as a traffic source: Chartbeat puts Google Search referrals to publishers down 40.2% year on year, while dark social and direct traffic grew. At the same time, AI answers are spreading across more assistants, not fewer. Comscore shows ChatGPT's share of AI prompts falling from 70% to 50% in six months as Gemini and Claude gain, and AI Overviews now appear on nearly four in ten US desktop searches. The offline evidence points the same way as last week's creative lesson: people say outdoor advertising feels more human than online ads, and the brands refreshing their identity this week are leaning on sound, belonging and heritage rather than product claims. For a business showing up in search, the condition is clear: earn the citation across several assistants, own the audience directly, and give people a real-world signal they can trust.
[ AI search ]
Google's September spam update hit hard over the weekend, with drops "across verticals and countries" and AI-generated content sites among the losers
The first stage of the update landed Friday, September 25 through Sunday, September 27, with SEMrush, AccuRanker, Mangools, Wincher, Zutrix and other trackers all showing heavy volatility. Glenn Gabe reported big drops across many sites, and AI-generated content sites, Grokipedia among them, lost visibility. The rollout is less than five days into a two-week window, so more movement is expected. Separately, Google is testing a "Loading..." button in place of "Show more" under AI Overviews, after an earlier "Enter AI Mode" label did not stick. (Search Engine Roundtable, Search Engine Roundtable)
Chartbeat: Google Search referrals to publishers fell 40.2% year on year, while dark social rose from 7.1% to 11.3% of traffic
Chartbeat's 2026 Publisher Playbook, reported September 24, measured a 40.2% drop in Google Search referrals from July 2025 to July 2026, nearly double the prior year's 21.9% decline. Google Discover fell 34.3%, up from a 6.6% decline a year earlier. Over two years, Google Search went from 9% to 5% of pageviews across the network. The channels that grew were dark social (texts, DMs, private shares), from 7.1% to 11.3%, and direct, from 13.5% to 15.9%. AI referrals grew eightfold but still made up about 0.01% of traffic. Loyal readers held steady while new readers fell 14%. (Search Engine Journal)
Comscore: ChatGPT's share of AI prompts fell from 70% to 50% in six months, Gemini rose to 30% and Claude to 11%, and AI Overviews now appear on 39.4% of US desktop searches
Comscore's Q2 2026 AI Intelligence Report, released September 22, found ChatGPT's share of total AI prompt volume dropped from 70% in January to 50% in June, while Gemini went from 17% to 30% and Claude from 2% to 11%. Google searches with an AI Overview reached 39.4% of US desktop searches in June, up from 25.8% in July 2025, and Bing searches with Copilot Search reached 17.3%. In June, desktop conversations totaled 168 million on ChatGPT, 89 million on Gemini and 22.3 million on Claude. (GlobeNewswire, Comscore)
Google Analytics can now allowlist the domains allowed to send data, and it drops everything else
On September 21, Google added an Include mode to GA4 hostname data filters. A property can list the domains permitted to send event data, and GA4 discards events from any other hostname, including events with an empty hostname, which usually signal spam. Measurement Protocol events are exempt. Filtered data is permanently removed from GA4 and BigQuery, processing takes 24 to 36 hours, and Google recommends running the filter in Testing mode first. (Search Engine Journal)
YouTube is building a Gemini editor that cuts Shorts by conversation, due in early 2027
Announced at Made on YouTube on September 23 and demonstrated on September 25, the tool builds a first draft from selected clips and a prompt, then revises it by chat: grouping clips by outfit color, cutting to a beat, adding shake and zoom, while keeping the manual timeline available. It arrives in early 2027 in Shorts and the YouTube Create app on Android and iPhone. YouTube did not specify languages, pricing or AI disclosure labels. (PPC Land)
Out-of-home feels "more human" than online ads to 69% of adults, and 87% say they would redeem a time-limited offer seen outdoors
New OAAA research with The Harris Poll, released September 17 and fielded August 4 to 5 among 2,015 US adults (analysis on 1,474 adults ages 18 to 64), found 73% prefer products or retailers that engage them through in-person experiences or local outdoor events, 69% say OOH feels more human than online or mobile advertising, and 64% say it feels more culturally relevant. On retail, 77% recalled seeing retail OOH recently (81% of Gen Z), 73% would visit a store the same day if OOH advertised product availability, and 87% would redeem time-limited OOH offers. (OAAA, GlobeNewswire, Billboard Insider)
Inca Kola rebranded to reclaim "the drink that brings Peruvians together," with a sonic logo made by tapping its own glass bottles
Design agency Epoch refreshed the Peruvian soda's identity, revealed September 27. Rival yellow-and-blue sodas, including Pepsi's Triple Kola, have been positioning themselves as more genuinely Peruvian since Coca-Cola bought into Inca Kola in the late 1990s, and a 2016 logo simplification that removed the outline of Peru drew criticism. Epoch kept the stripped-down logo but turned its shape into a frame for photography and video of the drink at social gatherings and with food, and added a four-note sonic logo, created with Echoic Audio, by tapping traditional glass cola bottles. (Creative Bloq)
Etam relaunched with a rounded logo, a new voice and 3,400 ad placements, made in-house
The French lingerie brand began rolling out its new identity on September 23: the capital "E" is kept but rounded, a new tone built on "passionate, confident and free," and imagery that pairs lingerie with boots, denim and everyday clothes, shot outdoors. Managing director Marie Delahaie, appointed this spring, called it an evolution in style rather than a break. The campaign runs across more than 3,400 placements, including bus shelters and Paris newsstands, plus a three-week online push in France, Benelux and Spain, and was created by Etam's in-house team. (FashionNetwork)
2026 is turning into the year of the pulled ad: Pepsi, Callaway, Novig and Oxford United all retreated within days
Creative Bloq's September 28 roundup lists the pattern: in July Pepsi deleted a Threads post for Wild Cherry within hours after it was read as a consent joke; in August Callaway pulled a 60-second parody ad the same day because, without the film reference, it looked like a man violently shoving a woman; in September the Novig ad with Sydney Sweeney drew criticism from female athletes, including Olympic swimmer Ariarne Titmus; and Oxford United trailed a "United 93" clothing range, marking its 1893 founding, for launch on September 11. (Creative Bloq)
The one-line read. Google is cleaning house and handing out links at the same time. The September 2026 spam update began rolling out on September 24 with a two-week window, while tracking data shows the share of AI Overviews carrying external links jumping from near zero to 26.2% in ten days. The catch is that a link is not a visit: some of those links open AI Mode rather than the source page, and a fresh click-through study shows the top organic result earning roughly half its usual clicks when an AI Overview sits above it. Paid AI inventory keeps spreading, with ChatGPT Ads now sold in more than 60 countries, and measurement vendors are turning their models into agents that move budget. The creative lesson of the week runs the other way: the most copied video of 2026 spread because it was so plainly made by people. For a business showing up in search, the condition is clear: be the cited source, measure the click and not just the mention, and keep proof of who is behind the work in plain view.
[ AI search ]
Google's September 2026 spam update started rolling out on September 24, and this one may take up to two weeks
Google announced the update around noon ET on September 24, saying it "applies globally and to all languages" and "may take up to two weeks to complete." It is the fourth spam update of 2026, after March, June and August, and the earlier three each finished in a couple of days. It is a general spam-policy update, not a link-spam update. It lands on top of the unconfirmed ranking swings of September 23 and 24. (Search Engine Roundtable)
AI Overviews with external links in the answer text jumped from near zero to 26.2% in ten days, but some links open AI Mode instead of a website
Peec AI's daily tracking, reported by Search Engine Journal, shows the share of AI Overviews with external links inside the answer text sitting near zero from August 25 to September 10, then climbing sharply from September 11 to 26.2% on September 20. Separately, on September 22, Search Engine Roundtable documented a test where underlined text in an AI Overview opened AI Mode rather than the cited page. SEJ notes Peec did not disclose sample size, markets or devices, so treat the number as directional. (Search Engine Journal, Search Engine Roundtable)
Advanced Web Ranking: the number-one organic result's click-through rate falls from 20.02% to 9.69% when an AI Overview is present
AWR's 2026 CTR study, built on more than 1.75 million US results pages a month, found AI Overviews on 43% of sampled desktop results in July 2026. Results pages with an AI Overview ended without an organic click 74.3% of the time, against 53.6% without one. Stacking hurts most: AI Overview plus People Also Ask plus videos cut position one's clicks by 81%. Navigational queries lost about 30 percentage points, while transactional queries were relatively protected. (Advanced Web Ranking)
Ann Smarty's second in-house survey: 39% of companies call it "AEO," 37% call it "GEO"
Ann Smarty's second "State of SEO for AI" survey of in-house teams and business owners, published September 23, found the two labels almost evenly split. The summary did not publish a sample size; the full report is on her site. Her qualitative read: fundamentals still carry the load, with Reddit, schema and content chunking as the added tactics. (Ann Smarty)
ChatGPT Ads expanded to seven more Asian markets on September 23 and is now sold in more than 60 countries
OpenAI added Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan. On September 1 it said the ads business passed a $1 billion annualized run rate in under 200 days with "tens of thousands" of advertisers, and that small and midsize businesses "account for a material share" of revenue. The platform now offers CPC and outcome-optimized bidding, a pixel and Conversions API, and more than 50 measurement and tech partners. (OpenAI, AI News)
Haus launched Architect, an agent that turns causal marketing mix modeling into budget moves, claiming 10.5% pooled lift in pilots
Announced September 21, Architect sits on Haus's causal MMM and recommends the "next best action for every dollar," simulating reallocations across channels while keeping a human approving each change. Haus reports a 10.5% pooled lift across budget recommendations in pilot engagements and cites SharkNinja, Dyson, FanDuel, Intuit and Wayfair as customers. It pitches against the finding that only 40% of marketers say their measurement tools let them act decisively. (PR Newswire)
News UK says a fifth of its website traffic over the past six months was not human
News UK chief technology officer Tom Jackson said 20% of traffic was non-human, with 6% actively blocked as malicious, and that bot traffic had been "significantly higher" before the company deployed Netacea protection. Press Gazette published the figures September 23. (Press Gazette)
The Online Responsibility Network launched a consumer "trust mark" for online brands, backed by 75% public support
Unveiled on the first Online Responsibility Day, September 24, with a giant London billboard comparing online content to food, the network introduced a trust mark to identify brands and platforms improving online safety. Its research found 80% of Brits support food-style warning labels on sites carrying misinformation and 75% back a trust mark. Founding members include giffgaff and Save the Children. (Famous Campaigns)
The most copied video of 2026 spread because it was visibly human-made, then became a template brands are cloning with AI
Romain Gavras's near eight-minute video for GENER8ION and Yung Lean's "STORM," released in April, climaxes in a 2.5-minute formation dance choreographed by Damien Jalet. InsideHook, on September 21, traced its spread: tech and VC figures shared it as proof of human artistry AI could not match, and then AI video tools such as Higgsfield let creators swap in their own faces and copy the moves. By September 24, brands across food, delivery, education, jewellery and technology were adapting the formation for their own ads. (InsideHook, Best Media Info)
The one-line read. Search is splitting into more doors, and the measurement is finally catching up. Google Search Console now reports image-led searches from Lens and Circle to Search as their own category, the UK regulator wants AI assistants on Android and Chrome choice screens, and freelancers report GEO requests rising faster than Google-ranking requests. At the same time, Google keeps taking small decisions out of owners' hands: a four-day window before user edits to a Business Profile can go live, hours that AI can set from a sentence, and budget modes that move ad spend between campaigns. The brand work that stood out this week did the opposite of automation: IKEA and WWF built campaigns from what real people already do and already value. For a business trying to show up in search, the read is to claim every surface Google now measures, and to guard the facts about the business as closely as the rankings.
[ AI search ]
Google Search Console added a "Multimodal" filter: searches started with an image, photo or screenshot now get their own report
Rolling out globally from September 24, the web search type in the Performance report now splits into "Text-based" and "Multimodal." Multimodal covers Google Lens, Circle to Search on Android, image uploads to Google Search, and Chrome's "Search this image." Google confirmed this is entirely new data: these searches were not previously counted in web results reporting. Query text is not available for multimodal rows, because the query is an image. (Search Engine Roundtable)
Rankings swung hard on September 23 and 24 across more than a dozen trackers, with publishers reporting 20% to 80% Discover and traffic drops
Search Engine Roundtable logged spikes on AccuRanker, Semrush, Mozcast, Sistrix, Similarweb and roughly ten other tools, following earlier instability on September 4 to 13 and September 15. Google has not confirmed an update. WebmasterWorld posters described Discover delivering a one-to-three-hour boost "and then flatlines," and one said September ad revenue "has literally been cut in half." (Search Engine Roundtable)
Google Business Profile now gives owners four days to reject a user's suggested edit, and is testing AI that sets hours from a typed sentence
Google's updated help page says owners get notified by email or in-product and have four days to accept or reject suggested edits; after that, Google may publish the edit if it finds corroborating information, such as the business website. Google also says it "may apply suggested user edits to your business information without prior review" in some cases. Separately, a test lets owners "Describe your opening hours" in plain text and Google's AI fills in the weekly schedule. (Search Engine Roundtable: four days, Search Engine Roundtable: auto hours)
The UK's CMA wants AI assistants on Google's Android and Chrome search choice screens, shown at first use and then every year
The Competition and Markets Authority proposed that Google offer a search-service choice when a user first sets up Android or Chrome and annually after that, include qualifying AI assistants on those screens, and make search providers properly attribute publisher content. Chief executive Sarah Cardell said AI assistants are "increasingly" used as people's search service but "aren't included on Google's choice screens." A final decision is due by October 9. (Search Engine Roundtable)
Fiverr: 67% of search freelancers saw GEO requests rise in the past year, against 56% for Google-ranking work, and 61% say clients expect AI to make it cheaper
Fiverr's 2026 Holiday Marketing Index, released September 21 and covering freelancers in eight countries, found 61% still rank Google SEO as the stronger long-term investment, 83% say clients now hire them more for advice than execution, and 61% put a typical project at $150 to $300. Fiverr says GEO services grew 29% from November 2025 to April 2026. Search Engine Journal notes Fiverr did not disclose the sample size or question wording, so treat the percentages as directional. (Search Engine Journal, GlobeNewswire)
Google Ads added "Holistic" and "Growth" investment modes that move or add budget across campaigns, with a 7-day forecast
On the Recommendations page, Holistic mode "reallocates existing spend from underutilized or less efficient campaigns" to top performers; Growth mode adds budget to top performers without cutting anything. Eligibility requires at least one budget-limited campaign or a forecast of significantly more conversions at a favorable CPA or ROAS, and estimates typically use a 7-day forecast. Excluded campaigns are left alone. (Search Engine Roundtable, Search Engine Roundtable: budget forecast)
Related measurement rule to act on: Google ignores offline conversions uploaded more than 7 days late when it calculates data-driven attribution
Google Ads now leaves conversions uploaded more than seven days after they occur out of data-driven attribution, although they still appear in standard reports. Two Octobers flagged it in its September roundup for advertisers who import offline conversions. (Two Octobers)
Google is testing bigger Local Service Ads with a larger image and a highlight box under the listing
Anthony Higman spotted the redesign, which expands the ad's visual footprint, and noted Google is investing heavily in the format while moving Local Service Ads into Google Ads. (Search Engine Roundtable)
WWF's "Where there's a tree..." won the Brand Impact Awards' Social Impact prize by selling what forests give instead of what deforestation takes
Johnson Banks built the WWF platform around an open-ended phrase, "where there's a tree...", that finishes with something the forest provides, replacing the usual doom messaging. A judge called it an "infinite storytelling device" that can explain complex ecology in a pithy way. Best of Show went to Fieldwork Facility's V&A East Storehouse identity, which treats a building holding more than 250,000 objects across 16,000 square meters as an "operating system," with wayfinding designed as a discovery tool. Results were published September 23: four Golds, 17 Silvers, 31 Bronzes. (Creative Bloq via Yahoo, Creative Bloq)
Apple is marketing its first foldable like a challenger: five days of explainer sketches on TikTok instead of a single cinematic ad
Apple's official TikTok ran a rapid series of short sketches for the iPhone Duo: the phone "doing yoga," acting as a pocket cinema, a cafe mini-laptop, and a ring box for a proposal. Creative Bloq notes it is unusual for Apple to explain this much, and reads it as Apple behaving like the challenger in a category Samsung defined. Attention is high: Marques Brownlee's Duo impressions video reached 26 million views and entered his top three in a week. (Creative Bloq, Max Frequency)
IKEA Italy turned the 30-year-old FRAKTA bag into a fashion-shot photo series of real customer hacks, built on a stat: 65% have given it another job
Ogilvy Italy's "Belli organizzati," launched September 19, is ten images by street photographer Salvatore Matarrazzo showing how Italians actually reuse the roughly one-euro blue bag: cat bed, toddler's paddling pool, record storage, laundry, moving day, instrument case. IKEA research found 65% of people who have owned or used FRAKTA have repurposed it. It extends IKEA's "Fai spazio a ciò che ami" (make space for what you love) organization platform. (Famous Campaigns, LBBOnline)
The one-line read. The web is quietly growing a second front door. Google's own auditing tool now scores whether AI agents can find a site's tools, a randomized study of 1,100 searchers found forced AI Mode pushed people toward rival search engines rather than deeper into Google, and review thresholds on Google surfaces are hardening into fixed numbers. On the brand side, the week's strongest work came from companies borrowing authority they already owned: McLaren's founder's gas-station sign, McDonald's fans' own social posts, JCPenney's customers standing in for models. For a business trying to show up in search, the practical read is that machine-readable proof and owned history are both becoming table stakes, and neither can be bought quickly.
[ AI search ]
Google Lighthouse 13.5 now audits whether AI agents can discover a site's tools, grouped with llms.txt under a new "Agent Discoverability" heading
Lighthouse 13.5 adds an audit for Agentic Resource Discovery (ARD), a proposed spec for how AI agents find the tools and services an organization offers. It checks for an Agentmap line in robots.txt, an `ai-catalog` link tag or HTTP Link header, then falls back to `/.well-known/ai-catalog.json`; no catalog means "Not Applicable," a broken one means fail. It sits in an experimental Agentic Browsing category, is expected in PageSpeed Insights within two weeks and in Chrome 156 DevTools, and already lags the spec: ARD v0.91 moved the manifest to `/.well-known/ard.json`, which Lighthouse does not yet read. (Search Engine Journal, DebugBear)
Randomized study: forcing AI Mode sent more people to Bing, DuckDuckGo and Yahoo, and hiding AI answers cost Google nothing in satisfaction
Researchers at the University of Pennsylvania and Northeastern ran a seven-day field experiment ("AI in Search Reduces Publisher Referrals Without Improving User Experience") with 1,444 enrolled users, 1,100 tracked and 956 surveyed, randomized into AI-hidden, normal Google, and forced AI Mode. AI Mode lowered satisfaction, trust and perceived relevance, significantly increased searches on competitor engines, and hurt heavy searchers more than light ones. Removing AI features produced no detectable drop in how users rated search. (Search Engine Journal)
Google Merchant Center put hard numbers on store ratings: 100 reviews, 3.5 stars, a rolling 24 months, calculated per country
Google added a requirements section: a merchant "generally needs" at least 100 eligible unique reviews, an average of at least 3.5 out of 5 for ratings to show on text and Shopping ads, and only reviews from the last 24 months count. Ratings are geo-segmented and calculated separately by country. (Search Engine Roundtable)
Google Business Profile now emails owners when it detects a spam-review spike, removes the reviews and pauses new ones for days
The notice reads that Google "detected a spike in spam reviews on your Business Profile and removed them so they do not affect your Google rating." New reviews and contributions are paused for a few days until the risk passes, and owners can request a manual look if real reviews were caught. One practitioner noted it can misfire on new locations getting a legitimate burst. Separately, the "Permanently closed" option has vanished from some profiles, with no Google statement yet. (Search Engine Roundtable: spam reviews, Search Engine Roundtable: closed option)
Parasite SEO found a new host: about 12,000 public Gemini Notebook pages ranking for porn, peptides and coupon codes
Black-hat SEOs published spam through public NotebookLM pages that Google indexed, roughly 12,000 of them, before Google pulled them from the index after Glenn Gabe, Gagan Ghotra and Charles Floate flagged it on September 21 and 22. (Search Engine Roundtable)
ChatGPT Ads' Shopify app goes international today, with HubSpot connected and OpenAI targeting $100 billion in ad revenue by 2030
OpenAI named Shopify its first ecommerce partner for ChatGPT Ads: the app syncs product catalogs, launches campaigns and tracks conversions, launched for US merchants on September 16 and opens to other ChatGPT Ads markets on September 23. HubSpot customers can now connect a ChatGPT Ads account. eMarketer puts OpenAI's ad target at $100 billion by 2030, forecasts 34.9% of the US population using ChatGPT in 2026 against 77.7% for Google Search, and says ChatGPT should "still be a small part of search ads budgets" for now. (eMarketer, Shopifreaks)
Snipp put brand rebates and loyalty programs inside ChatGPT and Claude, with no platform fees
Snipp Interactive launched a ChatGPT plugin and a Claude connector on September 14 that let shoppers discover and enroll in promotions, rebates and loyalty programs inside the conversation, with receipt OCR, SKU validation and fraud scoring behind one connection. Snipp lists no added platform fees or marketplace commissions. (PPC Land, Destination CRM)
McLaren unified Racing and Automotive under one wordmark drawn from the sign on Bruce McLaren's family service station
Announced September 22 ahead of the Azerbaijan Grand Prix, the identity replaces two systems with one. The new wordmark is based on the sign above the McLaren family service station in Remuera, Auckland, with a smaller, underlined "c" honoring the founder; papaya, the Speedmark and the Speedy Kiwi stay, inside an expanded palette tied to moments in company history. It spans Formula 1, IndyCar and, from 2027, the World Endurance Championship, alongside the push into performance SUVs. (PlanetF1, Ministry of Sport)
McDonald's UK turned fans' social-media envy into its biggest single-campaign menu launch: nine items from eight countries
"World Menu Heist" returns from Leo UK, launching September 22 with the first of two 30-second films. Social listening identified the international items UK fans most wanted, and those became the "targets": Germany's Big Rösti, Canada's Maple and Cheddar Chicken, China's Szechuan Sauce, plus items from the UAE, the Netherlands, India and Australia. The story unfolds in five heist stages (Plan, Execute, Plot Twist, Execute, Dispose) over eight weeks across TV, cinema, outdoor, social, radio and CRM. After its 2025 debut, more than 40 McDonald's markets ran their own version. (McDonald's UK newsroom, Famous Campaigns, Creative Salon)
JCPenney published "The Other September Issue" on Substack with five real teachers as the cover stars, eight looks for under $3,000 total
Debuting September 16 under the "Everyone Deserves a Better Deal" evolution of its "Yes, JCPenney" platform, the retailer launched its first Substack pop-up: a Vogue-style September issue starring five educators from San Marcos, Texas, styled by Wouri Vice, shot by Andrew Matusik and written by fashion critic Gabriella Karefa-Johnson. The teachers appear as fashion icons, not makeovers, and star in national TV spots that debuted September 17 during Prime Video's first football game of the season. All eight looks can be bought for under $3,000 combined. (Adweek, Marie Claire, Forbes)
The one-line read. Today the pattern is enclosure. Google is now testing links inside AI Overviews that look like links to a website but open AI Mode instead, Search campaigns are being auto-converted to AI Max without a rollback, Meta is removing the ability to exclude placements, and OpenAI has started refusing ads for products that compete with its own. Four different platforms, one direction: the surfaces a business used to steer are being converted into surfaces the platform steers, and the controls are coming out one at a time. The counterweight is unglamorous and it is the same in every case. Own the destination, own the list, own the proof, and make the content structurally easy for a machine to lift, because a citation is now the only durable form of referral and almost nobody has built for it.
[ AI search: AEO, GEO, SEO and local ]
Google is testing anchor links inside AI Overviews that open AI Mode instead of a website
Spotted on September 22 and reported by Search Engine Roundtable: links at the bottom of an AI Overview answer, styled like ordinary hyperlinks, carry follow-up questions and route the user into AI Mode rather than out to a page. Gagan Ghotra, who surfaced it, called it another tactic to push users from the classic results into AI Mode. There are no published metrics yet, only screenshots and video. Search Engine Roundtable
AI Mode on Gemini 3.8 Flash briefly stopped citing anything at all, and Google called it a bug
On September 3 the pattern was reported: AI Mode answers on Gemini 3.8 Flash were returning informational answers with no links and no citations. On September 4 Robby Stein, VP of Product for Google Search, replied publicly that it was not working as intended and a fix was coming; links reappeared the same morning. It sits inside a longer trend, not an isolated glitch. Between late April and the end of May 2026 the number of unique URLs cited by Google AI Mode fell 59%, and Gemini's citation usage fell 23 percentage points over a comparable window. Press Gazette's 2026 data puts Google referral traffic down 60% for small publishers, 47% for medium and 22% for large. Relevant Audience, Seer Interactive, Adapt Worldwide
44% of LLM citations come from the first 30% of a page, and cited pages use strict heading hierarchy at 68.7% against roughly 40% for uncited pages
Conductor's 2026 AEO/GEO benchmarks analyzed 13,770 cited domains across 3.5 million prompts, 17 million AI responses and more than 100 million citations. Alongside it, SparkToro's 2026 research puts 44% of LLM citations in the first 30% of page content, and Seer Interactive and BrightEdge find 68.7% of AI-cited pages use strict heading hierarchy versus about 40% of uncited pages. The Princeton GEO study, 10,000 queries, found the biggest gains came from machine-extractable provenance: quotations, statistics and citations, each worth roughly 25% to 40% more visibility. Conductor, Orbit Media, Known & Cited
Local search is being rebuilt around profile engagement and AI-generated Q&A, and early 2026 volatility hit 9.5 out of 10
The local picture for 2026: AI-driven local packs are displacing traditional local results, the old Q&A section is being replaced by AI-generated answers assembled from the profile, the site and public sources, and Google now lets reviewers display a different name and image than their profile carries, which makes reviews effectively anonymous. Three algorithm updates landed in four weeks early in the year and pushed local volatility to 9.5 out of 10, the highest reading of the year. Generic location pages that swap only the city name are being actively devalued. Sterling Sky, SEO.com, Digital Applied
ChatGPT Ads hit a $1 billion annualized run rate in under 200 days, then OpenAI cut off advertisers competing with its own products
Two things happened in the same month. ChatGPT Ads reached $1 billion in annualized revenue run rate in fewer than 200 days from launch, with self-service Ads Manager now live across 31 European markets and tens of thousands of advertisers. And OpenAI stopped accepting ads for products that compete with its own image and audio tools: Adobe lost the ability to promote Firefly without warning, in the same window OpenAI shipped ChatGPT Images 2.5 and ChatGPT Live. OpenAI's own disclosed projection is $2.5 billion in ad revenue this year, $11 billion in 2027 and $100 billion by 2030. PYMNTS, OpenAI, eMarketer
Google is auto-migrating Search campaigns to AI Max through September with no rollback
Between September 1 and 30, Google is automatically upgrading Search campaigns that use campaign-level broad match or standalone automatically created assets into AI Max. Campaigns keep their IDs but gain broader matching, text customization and final URL expansion. Since August 3, new campaign-level broad match and legacy ACA cannot be created in the UI, Editor or API. Dynamic Search Ads follow later: voluntary notices in September 2026, reminders in January 2027, automatic migration February 1 to 28, 2027. Once the window opened for an eligible campaign there is no rollback. Search Engine Land, Improvado
Meta is removing placement exclusions, and ChatGPT and Claude can now connect directly to Meta ad accounts
Meta began rolling out the removal of ad-set-level placement, platform, device and OS exclusions on August 20, account by account, with no announced completion date; Sales and Leads campaigns are going first. Separately, Meta's Ads AI Connectors now let ChatGPT and Claude connect to a Meta ad account without developer setup, and anything an agent creates defaults to paused pending human approval. B2The7 weekly roundup, citing Social Media Today and Meta for Business
Google Search Console's generative AI performance reports are now worldwide, with impressions but still no clicks or queries
The Search Console generative AI performance report has finished rolling out to all sites worldwide as of August 31, 2026. It covers impressions, pages, countries, devices and dates across AI Overviews and AI Mode. It does not include clicks, CTR or query data, and Google has promised more metrics without committing to a date. The rollout also carries a toggle letting a site block its content from AI Overviews, AI Mode and AI Overviews in Discover. The report had a bug earlier in September. Search Engine Land, Google Search Central
78 of 100 B2B SaaS and AI companies scored "generic" on brand distinctiveness, and 91 scored zero on sound
Louis Grenier's State of B2B Brand Distinctiveness scored 100 B2B SaaS and AI companies across distinctive-asset categories on a 24-point scale. One company landed in the "ownable" band. 78 scored "generic," 19 "recognizable," and 80% of the field was generic or invisible. Ninety-one companies scored zero on music, eight scored 1, nobody scored 2 or 3, for an average of 0.08. Color averaged 2.12 out of 3 with 86% scoring 2 or better, while human identity averaged 0.73 with only 14% scoring 2 or better. AI-era companies averaged 10.88 against 10.58 for pre-AI firms, a difference of nothing. STFO
Mattel told Brandweek it is a franchise manager, not a toy company, and replaced the funnel with a flywheel
At Brandweek in Atlanta, September 15 to 17, Mattel's CMO described the company as a franchise manager rather than a toy maker, and talked about fans and fandoms rather than consumers. The wider theme from the floor was funnels out, consumer flywheels in: journeys that move a person from product to film to TV to digital gaming so they meet the brand at multiple touchpoints rather than progressing down a line. Brandweek 2026 also ran a first-time AI Accelerator Day. Adweek, Creativ Brief recap
BrewDog bought back its own reputation with "Under New Ownership," including a letter from the new owner in 70,000 printed copies of City AM
BrewDog, now owned by Tilray Brands, launched a multi-platform campaign on September 21 built on one message: this is a new era under new ownership, with fresh investment behind the beers, the bars and the people. The media mix is large-format outdoor, digital screens in high-traffic areas, and a letter from Tilray printed inside 70,000 copies of City AM, plus a convenience-store promotion running alongside it. SLTN, Talking Retail
Contrast case: Brand USA's "America the Beautiful" answers the objection instead of selling the destination
Brand USA's global tourism platform, running into the 2026 World Cup, the 250th anniversary and the Route 66 centennial across nine priority markets, has expanded past scenery advertising into initiatives explicitly aimed at traveler confidence, including a strand named "Get Facts. Get Going." that exists to dispel misperceptions about visiting the US rather than to showcase it. The Drum, TravelPulse
The one-line read. The proof a business shows is becoming the thing that gets it cited. A Bluefish and Bazaarvoice analysis of 23,592 product pages cited by ChatGPT, Google AI Overviews and Perplexity in the first week of September found that 59% carry 100 or more reviews, 92% are rated four stars or higher and 65% put a review summary at the top of the page, while Google's own agentic commerce update reported that merchant-written "conversational attributes" were used in AI Mode recommendations half the time in a lululemon test. At the same time the surfaces that used to send the click are turning into surfaces that keep it: Google began testing a "Dive deeper" button in Discover on September 18 that opens a machine-written topic overview instead of the publisher's story, dropped free product listings from the European Economic Area under the Digital Markets Act, and Pinterest turned its 80 billion monthly searches, most of them visual, into a bidded ad unit. The measurement layer moved in the other direction, toward the physical world: Infillion bought Foursquare to tie ad exposure to store visits and purchases, and Fingerprint published a public directory of AI bots so sites can tell an agent from a scraper. On the brand side, Sonos spent its first campaign under Colleen DeCourcy telling audiophiles to "forget the calibrated acoustics," and a $500 million prediction market out-attentioned a $20 billion one by making its spokesperson an equity partner and creative director. The condition for a business that wants to show up in search: the answer engines are choosing sources by the depth of proof on the page, and the platforms are keeping more of the resulting attention for themselves.
[ AI search: AEO, GEO, SEO and local ]
59% of product pages cited by AI carry 100 or more reviews, 92% are rated four stars or higher, and 65% put a review summary at the top
Bluefish analyzed 23,592 brand and e-commerce URLs and 237,804 citation instances across ChatGPT, Google AI Overviews and Perplexity between September 1 and 7, 2026, in CPG categories, and Bazaarvoice published the results September 17 alongside a partnership between the two companies. Of the AI-cited product detail pages, 59% have 100 or more reviews, 92% hold ratings of four stars or higher, and 65% display a review summary above the fold; Bazaarvoice-hosted reviews appeared on about a third of cited pages with reviews. A companion Bazaarvoice survey of more than 1,700 adults across the US, APAC and EMEA found 57% name authentic reviews and star ratings as the top factor in trusting an AI recommendation, ahead of brand reputation, price and the AI's own explanation; 57% say it is "very important" to know the AI drew on real customer reviews and photos; and 62% trust AI recommendations more than they did six months ago. The partnership routes Bluefish share-of-voice data into Bazaarvoice's Influenster sampling, Content Coach and an Authentic Discovery API that makes review content readable to AI crawlers.
Google is testing a "Dive deeper" button in Discover that opens a Google-written overview instead of the publisher's story
Robby Stein, Google's vice president of product for Search, announced on September 18 that Google is testing a "Dive deeper" button on Discover cards that opens "a short topic overview with prominent links to related stories, community reactions & original reporting." The test starts on video cards; Damien Andell of 1492.vision found four client-side flags covering both video and article cards with "expand" and "inline" entry points, so the article version is built. The overview sits on top of the AI-generated preview summaries Discover already shows, with truncated publisher headlines and small thumbnails below it, and carries Google's standard "can make mistakes" label. Nicola Agius, SEO director at Reach plc, wrote that "Google chooses the angle" and publishers get no control over presentation. PPC Land's context: Discover drove two-thirds of Google referrals to 2,000 news sites as of August 2025, Discover page views fell 15% year over year in May 2026, Reach reported Google referral volume down 55% annually in July, and Ahrefs found in February that AI summaries in Search correlate with a 58% reduction in click-through. Google has said only that it will test "multiple designs."
Google dropped free product listings across the European Economic Area, with 90% to 100% visibility declines, under the Digital Markets Act
Search Engine Roundtable reported September 18 that Google Search removed free product listings and the organic product carousel for commercial queries across EEA countries including Germany, France, Belgium, Sweden and the Netherlands in mid-September, with tracking data showing roughly 90% to 100% drops. The change follows European Commission requirements under the Digital Markets Act; Google has previously described its DMA compliance changes as the biggest reduction in its search quality it has ever made. The space formerly occupied by free listings now goes to comparison shopping services and vertical search services, so merchants that relied on free Shopping visibility in those markets now reach the same shoppers through paid Shopping ads or a CSS partner.
Google's holiday agentic commerce update: merchant-written "conversational attributes" showed up in AI Mode recommendations 50% of the time, and a Business Agent now answers questions inside YouTube ads
Follow-up to the September 10 item on AI Performance Insights. At its Rethink ROI event on September 16 Google announced that AI Performance Insights in Merchant Center is now generally available in the US, Canada, Australia, India and New Zealand, showing a merchant's share of voice against competitors across AI Mode and AI Overviews. New "conversational attributes" in the product feed let a merchant supply answers to common product questions, compatible accessories and substitutes; in a test with lululemon those attributes were incorporated "50% of the time in relevant product recommendations in AI Mode." A Business Agent beta, by invitation for eligible US retailers, embeds inside YouTube ads so a viewer can ask product questions and get answers without leaving YouTube. The Universal Commerce Protocol integration hub is rolling out in the US with cart transfer to merchant sites and checkout-flow testing, with Australia and Canada in early 2027, and a video link attribute now auto-generates shoppable video formats. Google says merchants that adopt its core feed practices see a 5% monthly lift in conversions.
Pinterest launched Visual Search Ads: 80 billion searches a month, most of them visual, more than half commercial, now with a bidded top slot
At Pinterest Presents on September 17, Pinterest announced Visual Search Ads, a performance format that places an advertiser prominently in search results and Pin close-ups when a user searches by image, compares options, or moves toward a purchase. Pinterest says users run more than 80 billion searches a month, the majority of them visual, and more than half of those are commercial. Advertisers bid for the top response and drive the click to their own site; Modern Retail reported it alongside a wider suite of performance tools and an AI "restyle" feature. Search Engine Land's framing: the bet is that visual search unlocks intent based on what a user is looking at rather than what they type.
Infillion is buying Foursquare to tie ad exposure to store visits and purchases: 100 million points of interest, 16 billion check-ins, 250 million US devices
Infillion, the New York ad-tech company that owns Catalina's purchase data, announced September 18 that it will acquire Foursquare. Foursquare brings more than 100 million points of interest across 200 countries, 16 billion human-verified check-ins and aggregated coverage of 250 million US devices; combined with Catalina it gives advertisers a way to measure whether an ad drove a store visit and a purchase. Founder and executive chairman Rob Emrich told Axios the deal raises Infillion's revenue and headcount by roughly 20% to 25%; most of Foursquare's roughly 150 employees join Infillion's 650, and Foursquare keeps operating as an independent brand that sells data to customers on competing platforms.
Fingerprint published a public Bot Directory of AI tools and crawlers, with a Web Bot Auth compliance test for agents
Fingerprint, the Chicago device-intelligence company that says it identifies more than a billion unique devices a month for over 6,000 customers including Plaid, Booking.com, Dropbox and Trustpilot, launched Bot Directory on September 18: a continuously updated, searchable public registry of AI tools, bots and crawlers, listed by identity, provider and purpose. Operators can test their own agents for Web Bot Auth compliance and submit them for inclusion. Fingerprint's stated premise is that bots are now the majority of web traffic and that the useful question is no longer "human or bot" but "which bot, and what does it want."
Sonos told audiophiles to "forget the calibrated acoustics": the first campaign under Colleen DeCourcy repositions a speaker company as a feeling, two years after a $100 million app failure
"Love the Sound. Feel the System," Sonos's new global brand campaign, launched September 14 in the US and Mexico with a heavy New York out-of-home push: subway and station placements at Westfield World Trade Center, Moynihan Train Hall and Penn Station, painted billboards in Williamsburg, wildpostings, street panels and digital projections, plus video, podcast, print and social. Made by the in-house creative team, it uses hand-drawn "dance card" diagrams that double as system-configuration guides and wall art, collage and craft marks, and a voice-over that says "Forget the calibrated acoustics and software-driven analytics." Two products launch with it, Ace Ultra headphones and the Beam Ultra soundbar. The context: a 2024 app update that Sonos estimated cost $100 million in lost revenue and that the Wall Street Journal put closer to $500 million once the stock is counted, after which CEO Patrick Spence resigned. DeCourcy, who joined in November 2025 from Snap and Wieden+Kennedy, set two priorities: "to appeal to a more mass audience, not just early adopters, and to reconnect with longtime customers by getting back to what we do best." Q3 fiscal 2026 revenue was $375 million, up 9%.
Novig out-attentioned Polymarket with Sydney Sweeney as an equity partner and creative director: 1.2 million Instagram likes against fewer than 55,000 for a $15 million LeBron deal
Novig, a sports-only prediction market valued at about $500 million that launched nationwide in August under a federal framework, ran its first national campaign, "Just Sports," with Sydney Sweeney nearly nude behind sports equipment, telling men to trade on games rather than on wars or politics; 21-plus only, across digital, social, online video and out-of-home through the NFL season. Fortune reported September 18 that Sweeney approached the company, took equity instead of a fee, and worked as creative director on music, camera angles, coverage and script revisions. The one-minute social film drew more than 1.2 million Instagram likes and tens of millions of views across X and Instagram; Polymarket, valued at $20 billion, ran LeBron James, Derek Jeter and Eli Manning and drew fewer than 55,000 Instagram likes while reportedly paying James $15 million a year. System1 testing, reported by The Drum September 16, split by gender: men came "close to a five-star score," women scored it "around one star," with responses dominated by contempt, disgust and anger. The Drum's John McCarthy argued a product carrying the disclaimer "you should not trade with money you cannot afford to lose" should not carry visuals this compelling. Fortune's caveat is the American Eagle precedent: last year's "Great Jeans" campaign produced a sales spike and no sustained brand growth, with the stock back at pre-campaign levels within a year. Strategist Shruti Saini's phrase: Sweeney lets a smaller platform "out-attention" a bigger one rather than outspend it.
Instacart cast Shrek and Scooby-Doo to sell an AI assistant, made in-house, debuted on Sunday Night Football
Instacart's "Turn any idea into an order" campaign, reported September 18, sells Clementine, its AI shopping assistant built on 1.6 billion lifetime orders. Four spots use scenes people already know: "Solo Dinner" shows Shrek eating swamp fare while a viewer uses Clementine to find butternut squash risotto; "The Ultimate Sandwich" borrows Scooby-Doo for the high-protein sandwich trend; spots built on The Office and A Charlie Brown Thanksgiving follow this fall. Brian McGinn (The Bear, Chef's Table) directed; Instacart's in-house agency Local Produce created it and Prettybird produced; Hormel and Hillshire appear as CPG partners. It runs on linear TV, OTT, YouTube, audio, social and in-app, and debuted during NBC's Sunday Night Football. Jasmine Taylor, VP of brand marketing: "The more you use the product, the more it gets to know your preferences and routines."
Campari updated a 1970s ad for Negroni Week: "9 out of every 10,000 Americans prefer a Campari Negroni," read aloud by Alan Cumming
Campari's "Stay Bitter" campaign launched September 21 for Negroni Week (September 21 to 27). The idea is a single word added to a 1970s print ad: the original line "9 out of every 10,000 Americans prefer a Campari," which sold exclusivity by admitting near-total unpopularity, becomes "9 out of every 10,000 Americans prefer a Campari Negroni." In the hero film Alan Cumming reads "Bitter" magazine, finds the updated ad, builds a Negroni, and calls one made without Campari "a traitor," a nod to his hosting of The Traitors on Peacock. M Booth handled creative, Palette Group produced, and Cumming leads "The Bitter Nine," a tastemaker roster including chef Jonah Reider and fashion figure Jalil Johnson, in creator-led content; drinkers in select markets get up to $15 back on a Negroni receipt. Allison Varone, Campari America's head of marketing: "Consumers, young consumers, millennials, Gen Z, everyone, are feeling this nostalgia."
The one-line read. The people who recommend a product are becoming the raw material AI answers are built from, and the platforms are racing to package them. The IAB's Creator Signal study, published September 18, found that 56% of consumers who shop with AI tools prefer a recommendation that cites a creator or influencer review, rising to 65% among Gen Z and millennials, while 80% say they are comfortable relying on AI for product recommendations at all. On the same day Meta turned its creator tools into a single Creator Marketing Hub with an MCP connector so AI agents can build partnership ads, and OpenAI's European ChatGPT Ads rollout got its first named advertisers, Volkswagen and Vodafone, through Adform. Underneath the recommendation layer, the measurement layer got shakier: Google's September 13 change to how it serves results cut one rank tracker's data collection by about 80% and slowed several others, and an AI Mode button is being tested in the search bar on results pages, so a follow-up search can become a chat by default. On the brand side, YouGov's 2026 US Most Recommended Brands ranking put Zara first at 90.5% and Vrbo as the biggest gainer at 9.2 points, Staples inverted its own Easy Button into a black Hard Button, and Ben Cohen handed Brandweek a $20,000 guerrilla brief to free Ben & Jerry's from Magnum. For a business that wants to show up in search, the condition is this: the recommendation is the unit of visibility, AI systems weigh who made it, and the tools that used to measure the result are the ones being locked out.
[ AI search: AEO, GEO, SEO and local ]
IAB: 56% of AI shoppers prefer a recommendation that cites a creator, and everyday customers beat professional reviewers as the preferred source
The IAB's "The Creator Signal: How AI Systems Interpret Influence in Commerce," reported by Marketing Dive on September 18, surveyed 2,200 people aged 16 and over who had used AI tools for shopping in the previous three months, across the US, UK, Australia, Mexico and India, fielded in May 2026 on the Attest platform. 80% said they are comfortable relying on AI for product recommendations. 56% favor recommendations that cite a creator or influencer review; the figure is 65% for Gen Z and millennials, 53% for Gen X, and only a third of baby boomers, 66% of whom prefer other sources. 29% of Gen Z and millennials say a creator citation makes them more likely to buy, against 12% of boomers. 38% found influencers most helpful for finding the best price, and 11% found creator content unhelpful. Asked which kind of creator they prefer, more than 50% chose everyday consumers and 43% chose professional reviewers, and across every generation subject-matter experts, retailer information and brand content still outrank creator reviews for the purchase decision itself. IAB's Jack Koch: "Consumers are looking for authenticity and trust and that is coming from creators."
Google changed how it serves results on September 13 and cut one rank tracker's data collection by about 80%
Search Engine Roundtable reported September 18 that Google appears to be more successful at blocking scrapers and tracking tools following a change around September 13, which landed in the same window as the ranking reversals of September 4 and 12 and the September 15 update. Nozzle reported roughly an 80% drop in data collection from Google Search, DataForSEO reported reduced scraping success, and Sistrix has been running at a "reduced rate" since September 16, with a status note saying "Google has made further changes to the way search results are delivered." Ahrefs is still collecting normally and Semrush is slower but working. Glenn Gabe documented cases where different platforms showed different recovery timelines for the same ranking change, which the article attributes to varying scraping success. Barry Schwartz's warning: "just be careful when looking at the data from these tools, they might be wrong."
Google is testing an AI Mode button in the search bar on results pages, and opened its aggregator and supplier units to local business queries
Dwayne Cubbins spotted, on September 15, an AI Mode button inside the search bar on Google's results page rather than only on the home page, and Search Engine Roundtable confirmed it September 18. The button does not appear on every query, and where it appears a regular search still works, but Cubbins noted "now I have to hit Enter to continue with the regular search" and asked whether Google is "trying to force people to use AI Mode when making a follow-up search." Separately, Google updated its aggregator unit and supplier unit developer documentation on September 18 to state that both now support local business queries, pointing local businesses to the Local Point of Interest Feed documentation.
Meta consolidated creator tools into a Creator Marketing Hub, with an MCP connector so AI agents can build partnership ads, as the IAB puts US creator ad spend at $44 billion
Meta announced the global launch of its Creator Marketing Hub on September 17 at the IAB's first Global Creator Week. Facebook creators now join Instagram creators in the Creator Marketplace API, so an advertiser can discover both through one integration; a new Content Discovery API recommends top-performing organic creator content; Meta's AI business assistant is built into the hub to search for creators and recommend partnerships; and there are new ad-readiness editing tools and in-hub partnership messaging. Live video ads arrive on Instagram on September 29, and the Meta Ads MCP connector now supports creating and managing partnership ads from AI agents. Features roll out through the rest of 2026. Context from the IAB: creator economy spending reached $37 billion in 2025, 45% of consumer purchases are influenced by creator recommendations, and US creator ad spend is projected to reach $44 billion in 2026, growing four times faster than the media industry overall, up from $13.9 billion in 2021. Meta's Q2 2026 ad revenue grew 27% year over year to $60.8 billion on 14% more impressions.
ChatGPT Ads in Europe got its first named advertisers through Adform, and Wayfair was the first brand seen running a Sponsored Agent
Marketing Tech News reported September 17 that Adform is a technology partner for the European rollout of ChatGPT Ads, with Volkswagen testing brand presence in the conversational environment and Vodafone testing channel performance across the customer journey. OpenAI expanded ChatGPT Ads to 31 European markets in August; the platform now offers conversion-optimized bidding, geo-targeting and custom audiences beyond the original CPM and CPC options, with measurement through the OpenAI Pixel, a Conversions API and third-party integrations. Adform's Katja Henneveld said the format "creates a new moment for advertisers to be relevant." Separately, MarketScale reported September 18, citing Ad Age's September 14 reporting, that Wayfair was among the first brands spotted using the Sponsored Agents format announced this week; OpenAI projects $2.4 billion in ad revenue for 2026 with ads live in seven countries and 50 brands in the India rollout, while eMarketer forecasts US chatbot ad revenue will stay below $1 billion this year. Advertisers receive aggregated impressions and clicks; the conversation itself stays inside OpenAI's systems.
Staples built a black Hard Button as the antithesis of the Easy Button, and calls it a platform, not a campaign
Marketing Dive reported September 17 that Staples, under CMO Bob Sherwin (joined June 2025), launched a campaign around a black "Hard Button," the opposite of the brand's red Easy Button. Pressing it triggers workplace mishaps, such as restrooms running out of supplies or a print order reading "Golf Sharts" instead of "Shirts," each one a problem Staples' B2B services (furniture consultations, break room stocking, janitorial supplies) prevent. Created in-house with Bandits & Friends and produced by Monks, it runs on YouTube, Hulu, Netflix, LinkedIn and Meta, plus billboards, digital signage and elevator screens in corporate buildings, with shorter social cuts feeding mid-funnel display. Staples works with about 75% of the Fortune 500 but Sherwin says the brand has "a perception problem" and its business side is "one of the best kept secrets in business." His framing: "It's a platform more than just a campaign," because it can tell many workplace stories.
YouGov's 2026 US Most Recommended Brands: Zara at 90.5%, Vrbo the biggest gainer at plus 9.2 points, and 63% of Americans shared a brand experience in the last three months
YouGov published its 2026 US Most Recommended Brands ranking September 17, based on BrandIndex data from July 1, 2025 to June 30, 2026, with a minimum sample of 300 and six months of data per brand; the score is the share of current customers who would recommend the brand to a friend or colleague. Zara led at 90.5%, then Lush 90.3%, Savage X Fenty 90.1%, PUBG: Battlegrounds 89.5% and HOKA 89.5%. The biggest year-over-year gains were Vrbo, up 9.2 points from 69.2% to 78.4%, WE tv up 8.3 and Live Nation up 7.7. Sector leaders included Toyota (89.2%), Samuel Adams (88.7%), In-N-Out (87.3%) and Bloomberg (89.4%). A companion survey of 1,507 US adults (June 9 to July 8) found 63% had shared a brand experience in the past three months, 49% positive and 21% negative; 46% shared positive experiences to help others discover products and 52% shared negative ones to warn others.
Ben Cohen set a $20,000 guerrilla brief at Brandweek to pressure Magnum into selling Ben & Jerry's, with free ice cream for life as the fee
At Brandweek in Atlanta on September 16, Ben & Jerry's co-founder Ben Cohen asked the room to develop a "low-cost, quick, and easy-to-execute guerrilla marketing campaign" to pressure Magnum, which took over the brand in the Unilever spinoff in late 2025, into selling it. Submissions are due October 7 at 11:59pm, campaigns can be live stunts or social initiatives in the US or Amsterdam (Magnum's headquarters), the budget cap is $20,000, and the winner gets free ice cream for life if independence is restored. The backdrop: Magnum replaced CEO Dave Stever with a company executive, Jerry Greenfield resigned, the Free Ben & Jerry's petition has 200,000 signatures and about 800,000 social followers, and the brand's foundation faces closure at year end. Cohen: "The fate of Ben & Jerry's is in your hands. We need to pressure Magnum to sell."
Coach put Spotify inside its fall campaign and sent a Tabby Tour bus to college campuses on three continents
Coach's Fall 2026 campaign, "Live Your Story," launched September 16 with a global Spotify partnership. Four ads pair ambassador Elle Fanning with rising musicians PinkPantheress, Soyeon, Lilas and Shan Yichun. Inside Spotify, a "Live Your Story" hub serves playlists and lets users share a personalized listening persona. The Tabby Tour, a roving bus named for the brand's Tabby bag, kicked off September 16 in New York and runs through early 2027 across US, UK, Japanese and South Korean campuses, with performances and conversations from Tinashe, Rachel Chinouriri, Nxdia, Fousheé and Alessi Rose at select stops, tracked on an interactive map in Spotify. Visitors customize Coach hangtags, use a photo booth and pick up collectible merchandise.
BBH London opened a fake Angela Hartnett restaurant and served diners real school lunches
Famous Campaigns reported September 17 that BBH London, for the School Food Project (a coalition of Bite Back, Chefs in Schools, School Food Matters, The Food Foundation and the Jamie Oliver Group), launched Centena, a pop-up presented as chef Angela Hartnett's new restaurant. Diners were served authentic poor-quality school meals, processed chicken burgers on stale buns, soggy chips and cold baked beans, their reactions were filmed, and then the improved meals the project is campaigning for were served. Centena is an anagram of "canteen." The campaign launched alongside new UK government school food standards, which take effect in September 2027 for primary schools with secondary requirements phased in from September 2028. Chefs in Schools CEO Naomi Duncan: "This is a huge moment for school food."
Desigual cast Vivian Jenna Wilson opposite a robot that will not obey, and let the casting carry the message
Desigual's "Born to Disobey" film, reported September 16, pairs Vivian Jenna Wilson, Elon Musk's daughter, with a malfunctioning humanoid robot called DESI84 whose design mirrors Tesla's Optimus. Wilson's character walks it on a leash, uses it as a golf tee and tries to teach it to pronounce words; her line is "He was designed to obey. Not me." Directed by PSG, produced by ProdCo x Contrario, with creative direction from Desigual's Alexandra Sanhueza. Wilson has publicly criticized AI's effect on creative work, which the campaign leans on rather than hides.
Wonderful Pistachios launched its first NIL roster as trading cards, seven athletes across six sports
Wonderful Pistachios launched "Don't Hold Back All-Stars" on September 17, its first name, image and likeness roster: seven college athletes across six sports and Power Four conferences, including Maryland sprinter Quincy Wilson, Duke point guard Deron Rippey Jr., Colorado quarterback Julian Lewis, USC's Jazzy Davidson and Florida softball outfielder Taylor Shumaker. The in-house Wonderful Agency built a trading-card visual world with sport-specific creative per athlete, focused on how the snack fits each athlete's game-day routine, distributed on social across the fall, winter and spring seasons. Wonderful Agency's Bobby Pearce: "It needed to feel youthful and graphic." The better-for-you snack category is projected to grow from $118.1 billion in 2026 to $207.5 billion by 2034.
The one-line read. The ad unit and the business listing both turned into conversations this week. OpenAI made Sponsored Agents official on September 16: a ChatGPT ad now opens a labeled chat with the advertiser's own agent, with HubSpot and Shopify plugged in and 7, 14 or 30-day attribution windows to pick from. The next morning, Google was spotted rendering a local business's knowledge panel as an AI Overview with a "Show more" button that expands into an AI Mode chat, which means the model now writes the first description of a local business that a searcher reads. Underneath, the scoreboard stayed unfinished: John Mueller confirmed that Search Console counts an AI answer as one block rather than by link position and said the reporting will "evolve over time," while StatCounter's referral data put Perplexity at 4.31% and falling against Gemini at 10.9% and rising. Google also started showing advertisers what their peers spend, and the unsealed ad tech remedies gave publishers six years of court-supervised access to Google's auction. On the brand side, Gold House's Bing Chen told Brandweek that the monoculture is gone and the "expert economy" of credentialed niche leaders is what moves people, Crocs showed the receipts with $52.4 million in TikTok Shop sales and 200 items sold in a 12-minute live demo, and Cadbury launched Hershey's largest US campaign ever with five crystal clutches and a 30% increase in brand investment. For a business that wants to be found, the condition is this: the ad, the listing and the search result are all becoming a chat the business does not fully control, and the brands winning attention are the ones that show up as a credible person inside a small community rather than a logo in front of a large one.
[ AI search: AEO, GEO, SEO and local ]
Google is rendering the local knowledge panel as an AI Overview, with the model writing the business description
Search Engine Roundtable reported September 17 that Google is showing a local business knowledge panel with an "AI Overview" label at the top and a "Show more" button that expands into an AI Mode-style chat. Ben Fisher spotted it first on LinkedIn and Barry Schwartz replicated it for his own company, RustyBrick. The AI-written description drew on the business's own website; Schwartz noted that "most of it is correct, the niche solutions is a tad old" because the source page had not been updated recently. His reaction: "I am not sure I want AI describing my business, when I can." Google has been testing AI-generated descriptions for local profiles in several forms over the past year.
Search Console counts an AI answer as one block, not by link, and Mueller says the reporting will "evolve over time"
Responding to Kenichi Suzuki on September 16, John Mueller confirmed that citations inside AI Mode are not counted as sequential positions the way web results are; they are tracked as a block, as AI Overviews already are, and are not separated out in the Gen-AI performance report. Mueller: "The goal is not a written-in-stone absolute truth for position counting (that's impossible), but rather to make something that's useful for site owners," adding that "there will be some work required on your end when it comes to interpreting, understanding, comparing these metrics." He gave no timeline and did not say whether past data would be restated. For platform weighting, Greg Jarboe's September 9 analysis cited StatCounter data showing Perplexity's share of AI chatbot referrals falling from 7.91% in June to 4.31% in August while Gemini rose from 7.94% to 10.9%, and Similarweb's May figures putting ChatGPT at 53.9% of visits across seven assistants, Gemini at 27.9%, Claude at 9.2% and Perplexity at 1.3%. Siege Media's Ross Hudgens argued everyone should "remove Perplexity from their LLM trackers today"; Jarboe's counter was to de-weight it rather than drop it.
Google Search Profiles got a site badge, a redesign, and a lower follower bar
On September 16 Google published code and brand assets for a Search Profile badge that site owners can embed so visitors can follow the site's Search profile directly, with guidelines on touch-target size (48 by 48 dp on Android, 44 by 44 px on iOS and web), logo integrity, and pairing with the Preferred Sources badge. The same day Google redesigned how profile articles display, combined login management, and lowered the follower requirement for profiles. It sits alongside Preferred Sources and feed personalization as part of Google's push to let people follow specific publishers inside Search and Discover. Separately, Search Engine Roundtable noted September 17 that Google Discover on desktop, tested in 2023 and rolled out in 2025, has disappeared from Google.com.
OpenAI made Sponsored Agents official, with HubSpot and Shopify inside and attribution windows advertisers choose
Follow-up to yesterday's Digiday report: on September 16 OpenAI announced Sponsored Agents in ChatGPT Ads. After a relevant ad appears, a user can opt into a conversation with a business-sponsored agent that is "clearly labeled" and "distinct from ChatGPT's independent answers and separate from the original conversation," ask follow-up questions, and follow a link to the business's site when ready. It is in test with select US advertisers. OpenAI also added AI-assisted ad creation that drafts copy and imagery from the landing page and objective, opt-in text customization that adapts headlines to the conversation and translates them into the user's language, a conversion campaign type billed on impressions that optimizes for both click and view conversions, attribution windows of 7, 14 or 30 days for clicks and 0 or 1 day for views, and separate targeting for Android app, Android web, desktop web, iOS app and iOS web. HubSpot is the first CRM partner (create ads, track performance and manage leads inside HubSpot) and Shopify the first ecommerce partner, with the Shopify app going international on September 23.
Google Ads now shows advertisers what their peers spend, without saying who the peers are
Google rolled out a Spend Benchmarks report in the Google Ads account overview on September 16 that compares an account's weekly spend and clicks against businesses Google considers similar, based on industry and where the account advertises. One account shown was spending €284 a week against a €268 peer benchmark and getting 912 clicks against 765. Google has not published how many accounts make up a peer set, how "similar" is defined beyond industry and location, how often it refreshes, or whether advertisers can opt out of contributing. No Google Ads Help documentation had been located at publication.
The unsealed Google ad tech remedies: six years of a court monitor with source-code access, Prebid integration, and bid data for every publisher
AdExchanger published the unsealed remedies decision on September 16. Judge Brinkema, who declined on September 2 to force a sale of AdX, ordered behavioral remedies that take effect in 60 days and apply globally: a court-appointed technical monitor for six years with access to Google's employees, systems and source code; API integrations connecting AdX and DFP to Prebid; AdX bids submitted to rival publisher ad servers on the same terms DFP gets; real-time win and loss bid data shared with all publishers; published documentation of DFP's auction logic and pricing; and a prohibition on AdWords bidding directly into DFP or favoring Google's own tools. Integrations must be "functionally equivalent," so technical compliance with deliberate degradation will not suffice. DV360 faces no restrictions. Google can appeal; the parties file a jointly proposed final judgment by October 2.
Google Home will let any MCP-capable agent, Claude included, run the house
Google announced September 16 that Google Home is adding Model Context Protocol support, so any agent that can call MCP tools (Google named Antigravity, Claude, Hermes and OpenClaw among others) can interact with every device and the event history in a Google Home. Early access comes first to Google Home Premium Advanced subscribers in the US, with no launch date or pricing set. The Home app and Nest speakers still run on Gemini for Home; the swap applies to agents connecting through MCP, and it arrived alongside more than 15 other Home improvements.
Bing Chen at Brandweek: the monoculture is dead, the "expert economy" is what people follow, and mega-influencers are "too dilutive"
In a September 16 Brandweek conversation with Adweek editor in chief Ryan Joe, Gold House co-founder Bing Chen argued that culture is now "polyculture," hundreds or thousands of niches, and that brands should pick their core values and build conversations around those instead of commenting on everything. His four pillars for superfans: access to the brand, exclusive content, status designation, and experiences money cannot buy. He warned against "parachuting" into a community only to announce products ("you have to give before you receive"), called large influencers "too dilutive," and said the market is moving to an "expert economy" where the "credentialed leader within the niche" is who people follow. "The product is the customer," extended to the community of customers.
Crocs' TikTok Shop playbook: $52.4 million, a 744-hour livestream, and 200 items sold in a 12-minute demo
Crocs EVP and chief brand officer Terence Reilly used his Brandweek session on September 16 to demonstrate rather than describe. He had presenter Liana Gibbs host a live TikTok Shop stream from the stage; about 3,000 people watched and Crocs sold 200 items in 12 minutes. The context: Crocs joined TikTok Shop early in 2024 and did an estimated $52.4 million in sales in the 12 months to April, the top footwear seller on the platform per Charm.io, with more than 3 million TikTok followers. Its "Croctober" 31-day, 744-hour livestream marathon tripled its follower count, drew 550 million impressions and won a Shorty Award. Reilly's operating rules: "Do not build bureaucracy," keep teams and meetings small, and "give them permission to fail, and then give them permission to scale."
Cadbury's "There's Magic in More": Hershey's largest US campaign ever, launched with five crystal clutches at New York Fashion Week
Hershey launched "There's Magic in More," a multi-year platform and its largest US campaign for Cadbury to date, at LUAR's Spring/Summer 2027 show on September 14. The first activation was five limited-edition Cadbury Clutches designed with LUAR's Raul Lopez: jewel-purple, Swarovski crystals applied by hand, gold accents, and a platinum-silver insulated interior built to carry Cadbury XL bars and keep them cool. NYLON's editor in chief Lauren McCarthy and creator Tefi Pessoa carried them at the show. A 30-second hero spot directed by Tom Dream and an episodic social comedy series introduce the "Department of More," a fictional Cadbury unit where an intern learns how the chocolate is made, running across linear TV, streaming, digital video, paid social and display. A Pay It Forward sweepstakes awards the clutches to people nominated for generosity, with entries opening September 15. Hershey is raising Cadbury brand investment about 30% year over year on the back of 3.6% organic sales growth in Q2 and 5.8% over six months. Agencies: MiltonOne and Saatchi & Saatchi New York.
Norway's Consumer Council brought back the Enshittificator and gave him a partner: Death
The Norwegian Consumer Council released "The Life of Death and the Enshittificator" on September 17, a dark comedy from Oslo agency NewsLab, written and directed by August Jorfald, in which the professional product-ruiner from March's viral film (Bjørn Asgeirsson) teams up with Death (Frida Homlung): he gets products to the end of their lives, she takes care of the ending. The film accompanies a Council report on circular consumption, with the Council citing that two-thirds of consumers have discarded products because spare parts were unavailable and three-quarters have had electronics become unusable when software support ended. The asks: longer guarantees, cheaper repairs and stricter rules against designed obsolescence. Head of sustainability Elin Volder Rutle: "Quality must become the norm."
Fields Good launched a functional cookie on earned media, with the Mrs. Fields name as the credential
Ashley Fields, daughter of Mrs. Fields founder Debbi Fields, and co-founder Kim Anderson launched Fields Good in May, a cookie with functional claims such as focus and protein built around the argument that a healthy cookie should taste like a cookie. On the Brandweek stage September 16 with Adweek's Alison Weissbrot, Fields said the brand is "actually trying to shift people's perception of what a healthy cookie could taste like," and Adweek credited earned media as the engine of its early acclaim. No sales figures were disclosed.
The one-line read. The measurement problem in AI search got its clearest numbers yet this week, and they point in two directions at once. NP Digital's 68,000-answer fintech study found a challenger bank appearing in nearly half of relevant prompts and outranking Bank of America on first mention, because the models reward clear, consistent topical content over tenure and backlinks. In the same week, Tinuiti's data showed Amazon's citation share swinging from 4% on Copilot to 0.1% on ChatGPT to zero on Gemini, and 45% of brand decision-makers told the IAB that measurement is the single biggest obstacle to comparing AI journeys with traditional ones. So a small brand can win the answer, and nobody can yet say reliably which answer they won. The platforms kept moving underneath: OpenAI unveiled a ChatGPT ad that opens a branded chat instead of a website, Chrome started measuring how ad-heavy a page is, Google's rankings shook again on September 15, and Search Live got a new voice model. And while the digital channels get noisier, Lob's survey of 2,000 US adults found 87% of people read direct mail the same day it arrives and 62% now distrust digital ads more because of AI-generated content. For a business that wants to be found, the condition is this: the content rules now favor the clear over the large, the ad units are becoming conversations, the scoreboard is still being built, and the mailbox is quietly back in play.
[ AI search: AEO, GEO, SEO and local ]
A challenger bank beat Bank of America across 68,000 AI answers, and the same brand's citation share varies 40x by platform
NP Digital's Fintech AI Visibility Index, published September 14, analyzed more than 68,000 AI-generated answers to fintech prompts across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. SoFi appeared in nearly 50% of relevant prompts with an average first-mention position of 3.75, against 6.07 for Bank of America, and 61.8% of SoFi mentions were framed positively versus 46.8% for the incumbent. Gemini named an average of 10 brands per response against 6 for Perplexity, which VP Nikki Lam reads as Perplexity placements carrying more value because the list is shorter. The study credits relevance, topical consistency and clarity over brand tenure and backlink history. Digiday's measurement roundup the same week shows why single scores mislead: Tinuiti found Amazon's citation share at 4% on Copilot, 0.1% on ChatGPT and 0% on Gemini; Adobe and Semrush found Reddit at 28.9% of ChatGPT's citations while YouTube leads Google's at 21.1% with Facebook at 17%; and 45% of 200 brand and agency decision-makers surveyed by the IAB named measurement as their biggest challenge. Adobe's July survey of more than 5,000 consumers found 95% of AI users rate it as trustworthy as traditional search.
The follow-up query is the new keyword: map five kinds of second question, not one first one
Louisa Frahm's September 15 piece argues that conversational search interfaces, with follow-ups, multimodal input and memory, make the single-query keyword model obsolete. She cites Pew's finding that links inside AI summaries get clicked 8% of the time versus 15% when no summary appears, and that citation links inside summaries were clicked in only 1% of visits, and Google Lens handling 25 billion queries a month. Her framework maps follow-up paths around five types: clarification, constraints, comparison, validation and action. The tactics: build topic systems with hubs and supporting pages rather than 30 weak articles, state the answer first with verifiable evidence, treat visuals as answer assets, build "strong next turns" on owned surfaces such as calculators, comparisons and checklists, and earn "remembered authority" through distinctive datasets or original tools.
Search now confirms decisions made on YouTube and Reddit, and 84% of AI citations are earned, not owned
Karly Scott's September 15 analysis argues that the funnel has become passive exposure, preference, confirmation search, purchase, and that the preference-building step never shows up in attribution. She cites YouTube as the second most-cited social platform by AI systems after Reddit across more than 100 million citations, two-thirds of consumers rating YouTube "somewhat or very trustworthy" for purchase decisions, up to 70% of Gen Z following influencers with 44% having bought on a recommendation, and Muck Rack's analysis of 25 million AI responses finding 84% of citations come from earned media. Her prescriptions: build video libraries that answer product questions, keep product data feeds complete, participate in conversations you do not own, and stay consistent across the channels you do.
Google's rankings shook again on September 15, and Search Live got a new voice model the same day
Search Engine Roundtable reported a "huge spike in chatter" on the morning of September 15, with UK webmasters describing "mass movements," reports of spam on page one and AI Overviews showing "irrelevant nonsense," and Google Ads conversion drops coinciding with the organic volatility. Not all tracking tools had registered it at publication and Google has not confirmed an update; this follows the unconfirmed September 4 drop and September 13 recovery. Separately, Google released Gemini 3.8 Live and 3.8 Live Extended Thinking, with 3.8 Live now powering Search Live, the voice conversation mode in the Google app. VP Rajan Patel promised "more helpful responses, complete with web links," fluid multilingual support and more natural interaction, with transcripts and persistent history in AI Mode.
OpenAI's next ChatGPT ad opens a branded chat, not a website
Digiday reported September 14 that OpenAI is rolling out to select clients an ad format in which the call to action is "Chat with us": the user enters a branded chat interface inside ChatGPT and talks to the advertiser's own AI agent rather than leaving for a site. Wayfair is testing it; director of paid media Morgan Brown said the company is "testing at limited scale with guardrails around product accuracy, transparency and service handoffs." OpenAI CFO Sarah Friar told a Goldman Sachs conference on September 8 that everything so far is "before we've really launched a format that feels truly endemic to AI." Accuracast's Farhad Divecha compared it to Meta ads that send people to WhatsApp; Zeno Group's Shamsul Chowdhury said of ChatGPT ads to date that "the ROI just hasn't been there for the most part." No advertiser count, CPMs or test scale were disclosed; ChatGPT Ads reached a $1 billion annualized run rate in August.
Chrome now measures how ad-heavy a page is, and publishes it
Google added four experimental metrics to the Chrome User Experience Report on September 15: Ad Count (average ads visible in the viewport), Ad Density (average share of the viewport occupied by ads), Ad Weight Network (bytes consumed by ads) and Ad Weight CPU (milliseconds consumed by ads). Chrome identifies ads through network-level filtering and script analysis, samples the viewport once a second for count and density, accumulates network and CPU over the session, and reports at the 75th percentile. None are part of Core Web Vitals and Google has set no thresholds. WPP, The Guardian, Mediavine and Index Exchange endorsed the move as a way to connect cleaner experiences with inventory value.
Indig: 95% of companies get zero return on AI, and the cause is marketers building software
Kevin Indig's September 16 essay argues that marketing teams default to building AI automation in-house when they should first spend 30 minutes sorting work into buy, hire or build. He cites MIT's enterprise AI study finding 95% of organizations get zero return on $30 to $40 billion in generative AI spend, that bought AI tools reach deployment twice as often as in-house builds, that only 13% of marketers fully trust AI output without review, and that nearly 78% of C-suite executives have acted on an AI recommendation they later suspected was wrong. His rule: buy when the problem is common (rank tracking, citation monitoring, content scoring), hire when the workflow is proprietary but someone has built it before, and build only when automating one well-defined step with a clear input, output and verification. "You didn't hire your marketing team to write software."
Perplexity is hiring a creator-marketing lead with a multi-million budget, and paying for conversions rather than impressions
Digiday reported September 16 that Perplexity is recruiting an executive, at $175,000 to $225,000, to run creator marketing with a multi-million-dollar budget across more than 100 creators, from niche accounts to names like Lewis Hamilton and Joe Rogan. The role reviews creator "builds" made with Perplexity Computer to verify originality and real product use, and ties pay to conversions rather than reach. It follows OpenAI acquiring TBPN and hosting creator trips, one of which backfired, and Anthropic hiring an influencer-marketing lead. Analyst Ben Woods: "This is about getting potential users interested in AI's usefulness."
Direct mail is winning attention back: 87% read it the same day, 34% now trust it over digital, and 62% are more skeptical of digital ads because of AI "slop"
Lob's fourth annual State of Direct Mail: Consumer Insights report, released September 15 from a July 2026 survey of 2,000 US adults run with Kelsey White and PureSpectrum, finds 84% saying higher prices have changed their shopping habits and 68% paying closer attention to what lands in the mailbox. 82% collect mail daily or nearly daily, 87% read direct mail the same day, 79% took a digital action after receiving a piece, and 65% made a purchase, signed up or became a customer. 34% say direct mail is more trustworthy than digital, up nine points from 2025, and 62% say AI "slop" has made them more skeptical of digital ads. Among economically strained consumers, 55% say mail from a new brand makes them more likely to buy, against 32% of those less affected. Two cautions from the data: 45% discard mail using urgency tactics such as "Final Notice" or "Act Now," 56% say a genuinely useful offer matters more than a personalized name, and 67% are uncomfortable with AI-generated images of people even though 82% accept AI personalizing the offer. Lob CEO Ryan Ferrier: "Consumers are looking more carefully at where they spend their money, and that is changing what earns their attention in the mailbox."
Guinness found its new global asset inside a 1930s line; gFiber changed four colors and said nothing
Guinness and AMV BBDO's "LoVely" device replaces the V in "Lovely Day for a Guinness," a line with roots in the brand's 1930s advertising, with the harp, producing a flexible mark that debuted in Irish outdoor over the summer, is now integrated into Premier League activity and is rolling to Africa and Asia. Creative director George Hackforth-Jones: "The clever bit is resisting the temptation to invent another one," and instead finding "something new hiding inside an existing piece of Guinness language." The contrast arrived September 15: gFiber, Google's fiber business, announced a refreshed logo, typography and a teal, fuchsia, gold and orange palette across 19 states, with a release that said "a brand is more than a collection of colors and logos" and then explained none of the choices, leaning instead on three years of JD Power wins in the South region.
Bobbie cast the internet's from-scratch mother to sell the one thing she cannot make from scratch
Infant-formula challenger Bobbie, founded in 2018 with a formula free of corn syrup and palm oil, released a satirical thriller-style film starring Nara Smith, whose following was built on videos of cooking every meal from raw ingredients. In the spot she frantically searches parenting books and forums before concluding "there is no such thing as the perfect formula." The insight is that parenting perfectionism destabilizes people and that formula feeding needs destigmatizing; the casting makes the argument without saying it.
JCPenney publishes "The Other September Issue" today, with real teachers instead of models
JCPenney's "The Other September Issue" debuts September 16: a fashion-magazine-style publication in which real teachers model the retailer's affordable clothing, launched during Fashion Month as a counter to the celebrity-led luxury September issues. It is the next stage of the "Yes, JCPenney" platform launched in April 2025, under the line "Everyone Deserves a Better Deal." Marisa Thalberg, chief customer and marketing officer at parent Catalyst Brands, called it "the next evolution" of the platform. No results were disclosed at launch.
Apple TV floated a "definitely not cursed" billboard for a fictional town past Cape Cod's real ones
Apple TV sent a floating billboard promoting Widow's Bay, the fictional cursed-island setting of its comedy, around the waters off Cape Cod, near the Martha's Vineyard and Nantucket tourist traffic, extending the show's running joke that its mayor insists the town is "definitely not cursed." The series set a comedy record with 14 Emmy wins in a single year, and Apple has previously claimed it earned more Emmy nominations "than Martha's Vineyard, Cape Cod, and Nantucket combined."
KFC Australia turned a pickle into a fashion micro-trend with billboards for products that do not exist
KFC Australia's "Picklecore," from Special Australia, ran teaser billboards for imaginary pickle lipsticks, watches and trousers before revealing a 15-second high-fashion film with a techno soundtrack and a limited pickle menu (Pickle Burger, Pickle Pepsi Max, Frickles, Pickle Loaded Chips). Those That DO transformed a flagship restaurant with neon green light, pickle murals and a three-tier pickle chandelier. Special's co-CCOs Tom Martin and Julian Schreiber: "If Gorpcore could turn functional hiking gear into a high-fashion aesthetic, our Pickle Menu could do the exact same thing."
The one-line read. Two controlled experiments published this week put a number on what most AI-visibility advice has been guessing at: in a 30-day cold start, 85.8% of a brand's AI citations came from pages the brand did not own, and half of all cited sources stopped being cited within 30 days. The lesson is that AI answers are built from a brand's neighbors more than from its own site, and that the neighborhood turns over monthly. The platforms are moving on the same day. Google is testing sponsored text links written in the voice of the AI Mode answer, is letting Business Profile owners audit what its automated phone assistant has recorded about them, and is quietly pointing Performance Max clicks at a Google-hosted profile page instead of the advertiser's site. Cloudflare's new defaults, live today, block agent crawlers on any new ad-carrying domain unless the owner opts out. And in paid social, WordStream's 1,800-campaign benchmark shows real estate CPCs down 39.56% in a year. For a business that wants to be found, the condition is this: earned placement decides the answer, Google increasingly owns the landing page, and the cheapest attention right now is on Facebook.
[ AI search: AEO, GEO, SEO and local ]
Two GEO experiments: earned placements produced 85.8% of AI citations, and half of cited sources went dark within 30 days
Zeeshan Yaseen published results from two tracked experiments on September 14. In the first, a consultant brand was tracked on 15 commercial keywords across ChatGPT, Claude, Gemini and Perplexity over several months; it reached 37.01% peak visibility, listicles on third-party sites accounted for 72.4% of citations and PR for 24.1%, and a single Indeed SEO listicle generated 190 mentions, more than every other source combined. In the second, a SaaS agency with zero AI presence ran a 30-day cold start on 15 keywords across six platforms and logged 298 appearances, with 85.8% of citations coming from earned third-party pages and 14.0% from owned content. Three sources produced 78% of mentions. Adding competitor comparisons to the brand's own listicle increased its citations 12.25-fold. Half of all cited sources stopped being cited within 30 days. 18.5% of the agency's new users arrived via referral from AI platforms, and one Perplexity referral became a paying customer. The author revised his earlier advice: the owned listicle "is more of a foundation than a primary growth engine," and which sources mention you, and who appears alongside you, matters more than the content you control.
Google now lets Business Profile owners see, and delete, what its automated assistant collected by phone
Google launched a "Collected info" tab in the Business Profile dashboard on September 14, after a few months of testing. Google's automated assistant calls, texts or WhatsApps the verified number on a profile to confirm details such as hours and services, and answers given in those conversations can update the profile. The new tab shows each collected detail with its source and the date it was gathered, in its original language, and lets the owner delete anything inaccurate or outdated. Google's announcement on X: "Head to the 'Collected info' tab to see what was gathered and delete any outdated details." Separately on September 15, Natasha Kaurra spotted Performance Max quietly expanding a landing-page option that sends ad clicks to either the advertiser's website or a "Google-hosted page with built-in conversion measurement" built from the Business Profile.
Google is testing sponsored text links inside AI Mode answers, styled like the answer itself
Sachin Patel spotted, and Search Engine Roundtable reported on September 15, a Google test of ads in AI Mode formatted as text links with anchor text inside a response, on desktop and mobile, with a "Sponsored" label above the block. Google tested organic text anchor links in AI Mode responses in April 2026; this is the first sighting of them as paid placements. Barry Schwartz doubted the click-through rate, on the grounds that ads that look like the answer may underperform formats that look like ads.
A framework for AI Overviews and paid search: fight, influence, or generate demand
Ana Kostic's September 15 piece proposes sorting every paid keyword into three buckets. "Fight" is for commercial queries where clicks still convert (buy, quote, brand plus intent), where the rule is to pay for absolute top only if the brand is not already winning the answer box; her data shows ads appear in about 54% of queries costing $10 or more per click versus about 24% of queries under $2. "Influence" is for informational and mid-funnel queries resolved inside AI Overviews, where the goal is to be cited in how the answer gets built and to align ad copy with the Overview's answer rather than bid against it; in one B2B account she analyzed, 83.3% of terms triggering AI answers were informational. "Generate demand" is for zero-click queries where paid search cannot rebuild the funnel, and the budget moves to Demand Gen, YouTube and first-party content, with capture happening later through brand search.
Cloudflare's new defaults start today: agent and training crawlers are blocked on any new ad-carrying domain
From September 15, new domains onboarding to Cloudflare, new customers and existing Free-tier customers get new defaults: crawlers Cloudflare classifies as Training or Agent are blocked on pages that display ads, while Search crawlers stay allowed. Cloudflare defines Search as collecting content to answer questions later, Agent as automated behavior acting in real time on a person's behalf, and Training as taking content to train a model. Multi-purpose crawlers that combine Search and Training functions are affected by the Training block. Owners could opt out through the dashboard before today.
Google rebuilt Meridian with agentic model-building and brand signals, and took GeoX global
Reported September 14: Google's open-source marketing mix model Meridian now has agentic features that audit data quality, guide error resolution and steer model building in real time, plus back-end speed improvements. Marketers can feed brand signals such as Branded Google Query Volume into the model to show how TV, out-of-home and other upper-funnel spend moves brand demand and, eventually, sales. Meridian GeoX, Google's publisher-agnostic geographic incrementality tool, is now generally available after a beta. In the same measurement push, Data Manager is integrated into Google Analytics and Display & Video 360 and Google Ads gets a Data Strength Uplift Metric. Google describes the shift as turning measurement from a "reactive report card" into ongoing analysis.
Snipp put brand rebates and loyalty enrollment inside ChatGPT and Claude, with no platform fee
On September 14 Snipp Interactive released a ChatGPT plugin and a Claude connector, backed by an MCP server, that surface eligible brand offers when a user asks about a product and let the user enroll in a rebate or loyalty program without leaving the conversation. Purchase validation still happens by receipt upload afterward. The server exposes eight functions including receipt OCR, SKU-level validation, fraud scoring, an offer rules engine and audit logging. Snipp charges no additional platform fees or marketplace commissions on the AI surfaces. It cites 42% of consumers using AI shopping tools in the past month (a Morgan Stanley survey puts purchases via AI at about 23%), 86% of shoppers saying rewards influence purchases, and 28% abandoning loyalty programs that require another app. PPC Land notes there is no attribution between the in-chat offer and the eventual purchase, and that 55.9% of AI-driven traffic lacks attribution to its originating conversation.
WordStream's 2026 Facebook benchmarks: real estate CPC fell 39.56% and the cost of a real estate lead is $13.74
WordStream by LocaliQ's benchmark of nearly 1,800 Facebook ad campaigns, reported September 14, shows traffic campaigns at a 1.93% click-through rate (up 12.87% year over year) and $0.60 cost per click (down 14.29%). Lead campaigns run 2.70% CTR, $1.80 CPC, 8.54% conversion rate and $27.39 cost per lead (down 0.98%). Real estate had the steepest CPC improvement of any industry in traffic campaigns at minus 39.56%, and the second-lowest cost per lead at $13.74, behind only career and employment at $12.30. Dentists ($61.56) and beauty ($50.91) sit at the top.
Carter's, at 161 years old, replaced the apostrophe with a shooting star and moved from "cute" to "let every child's light shine"
Carter's announced a brand evolution on September 15. The new logo puts a shooting star in place of the apostrophe, with a refreshed script drawn from earlier marks in the company's 161-year history, and the brand promise becomes "Let every child's light shine." A 60-second film, "Watch Them Glow," runs through December 2026 on Nexxen, Roku, Disney+ and Hulu plus digital and social, and the company is launching a "Light Makers" creator program and partnerships with Outward Bound and Boys & Girls Clubs of America. Retail, product and packaging follow in 2027. CMO Sarah Crockett: "Our role isn't to create that light for them. It's to recognize it." The company operates more than 1,000 stores across the US, Canada and Mexico and carries a market cap of about $1.12 billion. This follows the 2024 "More Than Just Cute" campaign built on the finding that Gen Z parents let children pick their own clothes rather than dressing them to match the parent.
Jersey Mike's moved digital from under 1% of marketing spend to over 20% this year, and loyalty sign-ups rose 22%
CEO Charlie Morrison told Nation's Restaurant News on September 9 that before 2026 the 3,378-unit chain "really did not engage in any form of digital marketing," and that digital now takes more than 20% of the budget. Results so far: loyalty registrations up 22% year to date, ad awareness among Hispanic guests up 6% year over year, digital channels at 43% of sales (up 200 basis points), same-store sales tracking above 3% in the third quarter after 2.3% in the second. Brand awareness was already about 90%, so the digital push is aimed at frequency among Gen Z guests who know the brand but rarely visit, with call-to-action posts driving orders to the app. The TV layer continues: the NFL campaign with Danny DeVito and Eli Manning, from Highdive, positions DeVito as threatened by Manning's arrival after the chain became the league's official sub sandwich partner.
Kahlúa recruited three viral grandparents to turn a care home into a rave, then booked a real club night with a bring-your-grandparent door policy
Kahlúa Coffee Cream's "Not A Nightcap," from Wieden+Kennedy London and published September 15, is a 30-second film directed by Ciesay in which care-home residents swap Irish cream for Kahlúa Coffee Cream and turn the lounge into a dance party, starring grandparent creators Grimegran, Beau The Beard and Grandad Frank. The insight is that cream liqueur has been trapped in the evening ritual and the brand wants a daytime occasion. The film extends to a "Bring Your Own Grandparent" club night on October 8 at Number 90 in Hackney Wick with DJ Rossi, where guests who bring a grandparent get two free drinks.
Fake TfL safety notices on the London Underground mocked Meta's camera glasses, and nobody had to sign the work
Reported September 11: spoof stickers styled as official Transport for London safety notices appeared inside Underground carriages, in TfL blue with a simplified glasses icon and a Wi-Fi-style signal, warning wearers of camera-equipped smart glasses that "everyone thinks you look like a massive nonce." No creator is identified; the activist group Everyone Hates Elon has run similar stunts, and the stickers follow DuckDuckGo's "Normal F***ing Sunglasses" campaign and fake bus-stop ads earlier in the summer. The piece attributes the spread to growing discomfort about covert recording in public.
The one-line read. The week opens with the platforms starting to pay for what they take and the courts deciding who gets to take it. Google is now piloting cash payments to publishers whose pages materially shape answers in AI Mode, AI Overviews and Gemini, tracked in a new Search Console report, which turns "being cited" from a vanity metric into a line item. In the same window the Ninth Circuit ruled that an AI shopping agent acting through a user's browser is legally the user, not the AI company, which removes the main federal statute retailers were using to keep agents out. Google's own ads data shows why both matter: one-to-two-word queries fell from 42% to 24% of impressions since January 2025 while three-to-four-word queries now carry 46% of conversions, meaning the traffic that converts arrives as full questions. And in Australia the regulator forced the dominant real estate portal to drop its all-listing contracts, opening the door to listing-by-listing competition. For a business that wants to be found, the pattern is consistent: answers are becoming an asset with a price, agents are becoming legitimate visitors, and the walled gardens are being pried open one clause at a time.
[ AI search: AEO, GEO, SEO and local ]
Google is testing payments to publishers whose content shapes AI Mode, AI Overviews and Gemini answers
Google confirmed on September 14 an "early-stage learning pilot" called AI Contribution that pays selected publishers when their content "significantly" contributes to AI-generated responses across Gemini, AI Mode and AI Overviews. Participants get an AI contribution report inside Search Console that tracks when their pages influence an answer, with earnings accruing against that contribution. The report first surfaced in April 2026 without the payment element being clear. Google has approached at least dozens of publishers, the program has drawn more interest from small and mid-sized sites than from large ones, and it has expanded beyond news. Payment amounts are undisclosed.
Google Ads data: short queries lost 18 points of impression share, and three-to-four-word queries now carry 46% of conversions
Jason Tabeling's September 11 analysis of Google Ads account data from January 2025 through August 2026 shows a steady shift in query length since AI Mode arrived. Impression share for one-to-two-word queries fell from 42% to 24%, while three-to-four-word queries rose from 33% to 48%. Conversion share moved further: one-to-two-word queries dropped from 62% to 52%, three-to-four-word queries climbed from 20% to 46%, five-to-six-word queries tripled from 3% to 9%, and queries of seven or more words quadrupled from 1% to 4%. The trend accelerated each quarter.
Holland: the retainer model was built on the cost of writing, and that cost is now zero
Andrew Holland, SEO director at JBH, argues in a September 14 essay that AI made "the production of words incredibly cheap," collapsing the margin agencies built on content retainers, and that the scarce input is now the idea worth communicating. He cites organic clicks down 42% across 64 publisher sites since AI Overviews expanded, and uses The Cuckoo's Calling (roughly 1,500 hardbacks sold as Robert Galbraith, a bestseller once J.K. Rowling was revealed) to make the point that context and authority, not prose quality, decide what gets read. His prescription: writers move upstream to decide what the brand uniquely knows and what the market misunderstands, and content gets judged on economics rather than on whether a human typed it.
A September 4 ranking drop reversed on September 13, and Google never confirmed either move
Glenn Gabe reported that "a number of the sites that fell off a cliff around 9/4 completely surged back" on September 13, in both Search and Discover. Volatility trackers (AccuRanker, Mozcast, Semrush Sensor and others) had flagged September 3 to 4, and WebmasterWorld participants saw Discover traffic recover over the weekend. Google did not confirm an update in either direction, and the SEO community's reaction was unusually quiet.
The Ninth Circuit ruled an AI shopping agent acting through a user's browser is legally the user, not the AI company
PPC Land reported September 14 on the Ninth Circuit's decision vacating Amazon's preliminary injunction against Perplexity's Comet shopping agent. Comet captures screenshots in the user's browser, sends them to Perplexity's servers for interpretation, and executes navigation from the user's own machine, so "Perplexity's servers never directly touch Amazon's servers." The panel held that "the person at the keyboard" is the visitor, which means the Computer Fraud and Abuse Act's first element (intentional access to a protected computer by the defendant) is not met. The ruling leaves retailers with terms of service, rate limiting and technical blocking, but removes the federal computer-fraud statute as a way to keep user-directed agents out.
Australia's regulator forced the dominant real estate portal to drop its all-listing contracts
REA Group, operator of realestate.com.au, executed a legally binding three-year undertaking on September 10, accepted by the ACCC on September 11 and announced September 14. REA will immediately stop enforcing clauses that required agencies to list every property on the portal, stop tying training and marketing sponsorship (the Advantage+ program, since 2019) to top-tier depth contracts, and from January 1, 2027 let agencies move at least 25% of eligible sale and rental listings to lower paid tiers. The ACCC found the portfolio commitments prevented rival portals from competing "listing-by-listing" and may have substantially lessened competition, even though agencies were always free to list elsewhere too. REA says it "does not share the ACCC's concerns" and admits no wrongdoing.
Magnite and ITN are handing local TV buying to AI agents across 1,100 stations, in a $21 billion market with no cost data yet
Announced September 10 and reported September 12, ITN's seller agent now connects to Magnite's Orchestration layer, with execution through ClearLine and agent-to-agent communication over the Advertising Context Protocol, covering more than 1,100 broadcast stations across 75 ownership groups. A buyer submits one RFP covering local linear and connected TV together, and the companies say the cycle compresses from weeks to hours by removing phone calls, emails and manual make-goods. Full deployment is targeted for Q4 2026. US local TV advertising was a $21 billion market in 2025. No CPM, pricing or forecast-accuracy data was disclosed, and a June 2026 DataBeat study found conventional buyers still holding a 13.4% CPM advantage over AI agents.
Google Ads promotional credits are being invalidated after advertisers spent the qualifying amount
Search Engine Land reported September 11 that some Google Ads promotional credits are being marked "Invalidated" after advertisers have already spent the amount required to earn them. In one documented case an advertiser spent $3,200 to unlock a $3,200 credit, and the credit was revoked more than a month later. There is no known appeal path. Google Ads Liaison Ginny Marvin acknowledged the report ("I've passed this along to the team") without explaining the cause or committing to a change.
Columbia brought humor back on purpose, won its first Cannes Grand Prix, and added new customers at a double-digit rate
Columbia Sportswear's "Engineered for Whatever" platform, launched in 2024 with Adam&eveDDB, has produced "Expedition Impossible," which invited flat Earthers to find the edge of the planet with CEO Tim Boyle and won the Grand Prix in Brand Experience and Activation at Cannes Lions 2026, the company's first festival win, and a fax-only promotion that gave away 1990s Helvetia fleeces to customers who ordered by fax, with influencer Ken Eurich. Head of marketing Matt Sutton, in a September 14 interview: "Humor was kind of canceled out of advertising for a period of time, and we've intentionally brought it back." The company reports a double-digit rise in new North American customers in Q2 on net sales up 2% to $614.4 million, with US softness persisting. The platform deliberately reaches back to the brand's edgy 1980s and 90s "One Tough Mother" era.
Old Navy anchored back-to-school on one creator and got a 22% traffic lift on a declining business
Gap Inc.'s Old Navy built its entire back-to-school campaign around MrBeast, the first from the brand "anchored on a single creator," as the opening move in a multiyear deal. Results reported September 11: a 22% jump in online traffic to the kids' assortment after launch, more than 100 million YouTube views, and 1.4 billion impressions across platforms, with activations including a year of free lunch for a Kansas middle school and meme-trivia giveaways. Damon Berger, SVP of shared marketing services: "we believe that creators are the ones that driving culture more than almost anything out there." The shift comes after a 4% comparable-sales decline at Old Navy in Q2 and underperforming summer initiatives; MrBeast brings more than 515 million YouTube subscribers.
Duolingo took a fashion collection to NYFW where the price of admission was your streak, not money
Duolingo's in-house team, with designer Megan O'Cain, produced a seven-piece SS27 capsule for New York Fashion Week in which every piece carries the phonetic spelling of a luxury name (a lime-green bag reading "air-mess" for Hermès), sold at a two-day pop-up at Null Object where visitors bid using their Duolingo streak rather than cash. The insight came from YouGov research with Duolingo: 62% of Americans were confident they could pronounce Hermès and 15% got it right. CMO Manu Orssaud: there is "something very human about avoiding a word because you're worried you'll say it wrong."
Savage X Fenty hung giant lingerie over Seventh Avenue and let the sun do the media plan
On September 13, during New York Fashion Week, Savage X Fenty suspended oversized bra and underwear structures above Seventh Avenue and Fashion Avenue so their shadows fell across the road, the crossings and the pavement, shifting as the sun moved through the day. The creative changes with the conditions, and the audience was already treating the street between shows as an extension of the runway, so the installation generated content without asking anyone to post.
IHOP brought back an item it dropped six years ago because social listening said it was still the most-requested thing on the menu
IHOP relaunched stuffed French toast, absent for six years but still its most-requested item on social media, with a 30-second spot from Deutsch and DeVries Global styled after Sofia Coppola's "Marie Antoinette," soundtracked by Katseye's "Pinky Up," and positioned as an affordable luxury. The team went back and replied to years-old customer posts asking for the dish. Campaign metrics were not disclosed; parent Dine Brands reported domestic same-store sales up 1.5% in Q2 2026, outperforming benchmarks for a third straight quarter, and credited value-focused advertising.
The one-line read. Today's theme is that the shortcuts are being priced out. Reddit's share of ChatGPT citations fell from 3.83% to 0.5% in a matter of weeks, a survey of 1,042 content marketers found that 92.4% now use AI and it has no relationship with results, and Google's John Mueller said out loud that "position 1 to 10" no longer maps to anything useful. At the same time the ad layer is automating the tactical work away: Ritson's read on AI Max is that Google has removed the thing marketers were doing instead of marketing. What is left, and what every strong item today has in common, is strategy and craft that platforms cannot take from you: a positioning move (Autodesk, 7 Brew), an owned asset found inside your existing language (Guinness), or a product built to solve a fifteen-year-old customer pain (IKEA). For a small operator this is good news. The cheap tricks were never going to be yours anyway, and the expensive ones just stopped working for everyone else.
[ AI search: AEO, GEO, SEO and local ]
Reddit's share of ChatGPT citations collapsed from 3.83% to 0.5%, and citation sets now drift on a weekly clock
AdExchanger surveyed the AEO vendor field on September 10 (Profound, Envive, Trendos, Brandi AI, Optimizely, Noble, plus a Reddit spokesperson) on which tactics actually work. The headline data point: Reddit citations in ChatGPT dropped from 3.83% of citations in the July 18 to August 7 window to 0.5% by mid-August. Vendors reported that citation sets drift on weekly timescales, and Noble's publisher outreach model, which asks publishers to edit existing articles, gets a 4% to 10% response rate with reputational risk attached. The consensus on what holds: honest, human-written, structured content on owned sites, refreshed regularly, with strategy varying by category (apparel needs owned-site authority, finance leans on third-party sources).
Mueller: "position 1 to 10" is hard to map, and Search Console counts generative results as a block
Responding on Reddit to a question about how Search Console counts impressions for generative AI features, Google's John Mueller wrote that the old position 1 to 10 "is hard to map, or to make useful for site owners," because modern results pages offer many ways to interact. Search Engine Roundtable's read is that Google tracks AI results as a single block rather than assigning positions inside it.
Blogging success fell to a 12-year low, and using AI showed no relationship with results
Orbit Media's 2026 survey of 1,042 content marketers found only 14% report strong results from their blog, down from 26% in 2022. AI adoption reached 92.4%, and the survey found no relationship between using AI and stronger results. The average post now takes 3 hours 20 minutes, down from over 4 hours. What did correlate with results, and what fewer people are doing: keyword research, original research, formal human editing, paid promotion and consistent analytics. Influencer collaboration on content fell from 25% in 2017 to 7%.
Yelp put OpenAI's new voice model into its service-business phone agents, at $249 a month and 2.4 million annualized calls
The same day OpenAI released GPT-Live-1 in its API (September 10, priced at $0.05 per minute), Yelp deployed it inside Yelp Host, its restaurant phone agent, and Hatch, the home-services lead platform it bought in January for roughly $270 million plus $30 million in retention. The model swap enables full-duplex conversation: interruptions, overlapping speech, tone detection and automatic language switching. Yelp Host passed 1 million cumulative calls in June, ran at a 2.4 million annualized call rate in July, and has grown 38% month over month on average. Pricing starts at $249 a month after a 30-day trial. Hatch manages tens of millions of leads a year and was at $25 million ARR last November, growing 70%.
IAB raised its 2026 US ad growth forecast to 12.3%, and 76% of buyers say they are optimizing content for AI answers
IAB's September 10 update, from a survey of more than 200 brand and agency decision-makers, lifted the full-year US ad spend growth forecast to 12.3% from 9.5% in January. Social media leads at 16.5%, connected TV 15.6%, commerce media 13.6%, paid search 8.1%, linear TV at negative 1.5%. Customer acquisition is now the top priority for 63% of advertisers, up 9 points since January. On measurement, 86% expect AI to change their approach within 12 months, 48% name measuring brand visibility inside AI tools as their primary response, and 76% are increasing focus on optimizing content for AI-generated answers, even though only 16% systematically track AI visibility today.
Ritson on AI Max: Google automated the weakest part of marketing, and strategy matters again
Mark Ritson's September 7 column reads Google's September 1 migration of Search campaigns to AI Max, which now writes headlines and picks landing pages, as the removal of tactical work rather than a loss of control. His argument: platforms have taken away "the thing marketers were doing instead of marketing," which pushes the job back toward positioning, targeting, product, pricing and management. He cites Google's claim of 7% more conversions at similar CPA against an independent analysis finding roughly 35% lower ROAS versus traditional match types, Lunio's finding of 72% more invalid traffic on AI Max campaigns, and at least 18,000 roles cut across Omnicom, Dentsu and IPG in 18 months.
Autodesk's CMO runs marketing on "facts and feelings": build for the AI agents that now shortlist you, and for the humans who still choose
In a September 9 interview, Autodesk CMO Dara Treseder described a two-track playbook. On facts: "The agent is just following the facts," so the company builds its tech stack so AI agents evaluating vendors can find and verify what Autodesk does. On feelings: emotional work aimed at humans, including a 60-second Super Bowl spot built around a newborn and the 300 million people who design and make, and an LA28 partnership emphasizing zero new permanent venues. The positioning move underneath: repositioning from "the AutoCAD company" to a design-and-make platform under "Let There Be Anything." On creators, her rule is to find community voices and "let them cook" rather than script them.
7 Brew is the fastest-growing brand in the US, and its first CMO's strategy is to protect what was already working
7 Brew, the drive-through coffee chain, has passed 800 locations in 38 states and is on pace for more than 1,000 units in 2026, ranked fastest-growing brand in the US by Yelp. Nick Chavez, ex-KFC US CMO, joined as its first CMO in late 2024, and in a September 8 Q&A described the job as leaning into momentum rather than redirecting it: 685,000 TikTok followers and 7.5 million likes from in-house parody content (2001, Indiana Jones, Mean Girls), a loyalty app that hit number one on the iOS App Store, and physical stickers and coasters designed to travel. Amp is media agency of record.
Campbell's is putting 85% of working media into social, influencers, e-commerce and AI platforms
Campbell's, with net sales down 8% year over year in fiscal Q4 and snack organic sales down 6%, announced on September 8 that social, influencer, e-commerce and AI-powered platforms will make up roughly 85% of its working media budget, concentrated on growth brands Rao's, Goldfish and Pepperidge Farm rather than spread across the portfolio. Rao's consumption is up 9.4% with household penetration nearing 19%. The company is also cutting 13% of its workforce, closing two snack plants and targeting $500 million in savings.
Johnsonville ran a political attack ad for sausage, with AI backgrounds and a food-bank tie-in
Johnsonville launched "Sausage for Dinner" on September 10, a parody of midterm election advertising timed to the ad flood: red-and-blue creative, a "Sausage in Every Air Fryer" slogan, a 30-second TV spot, billboards, lawn signs and community events. The research backing it: 81% of Americans say society has become too political (Harris Poll), 31% say current events distracted them from figuring out dinner, and 53% dislike choosing what to make each day. Political ad spend in Grand Rapids alone is projected up 75% over 2024. The spot used generative AI for backgrounds to cut production cost, and donations go to Hunger Free America with election-themed swag for the first 100 donors.
IKEA solved Skyrim's fifteen-year-old inventory problem with a sentient KALLAX, voiced by Matt Berry
IKEA and agency Mother released a free Skyrim Creation on September 11 in which a new quest leads players to a Dwemer furniture store and introduces KALLAX Storageborn, a sentient shelving unit voiced by Matt Berry that offers unlimited storage, automatic loot collection and home tidying, summoned by the Dragon Shout "AYE KEY YAH." There is a clutter dungeon and a Clutter Golem boss fight. Mother's ECD Nick Hallbery: "It's been 15 years and no one had solved the biggest pain point for players." Distribution is through Bethesda's in-game Creations menu, backed by digital, social, in-store and PR.
Guinness found a new brand asset inside a 90-year-old slogan by swapping one letter for the harp
AMV BBDO's "LoVely" work for Guinness takes "Lovely Day for a Guinness," a line dating to the 1930s, and replaces the V with the harp, turning the slogan into a flexible visual system. It ran outdoor in Ireland first, moved to Premier League activity in the UK, and is expanding across Africa and Asia. Famous Campaigns frames it as part of a broader 2026 pattern of established brands finding assets they already own rather than inventing new ones.
The one-line read. Today's theme is the disappearance of the intermediate click. People Also Ask went from 86% AI-generated in August to 97% in early September, Apple shipped a Siri that answers web questions directly to 2.5 billion daily requests, and Google put AI search intent data inside Merchant Center so retailers can see the queries they never got a click from. At the same time the ad layer inside AI is hardening fast: Amazon plugged its DSP into ChatGPT, ChatGPT Ads crossed a $1 billion annualized run rate, and OpenAI quietly started refusing ads from competitors. The strategic read for a small operator is that the answer surface is consolidating faster than the ad surface is, so the cheap window is in being the cited entity, not the paying advertiser. And a leaked Google Geostore file this week says something useful about how to become that entity: it is not your listing, it is whether Google can assemble a complete, consistent picture of you across the whole web.
[ AI search: AEO, GEO, SEO and local ]
People Also Ask is now 97% AI-generated, up from 86% in a single month
Two independent datasets landed this week showing Google's People Also Ask boxes have been almost fully converted to AI-generated answers. AlsoAsked, working from 19.2 million English-language queries in 2026, measured 97% AI-generated by early September, up from 86% in August and from roughly 12% fourteen months ago. A second tracker, Allintitle's Also Ask Miner, already reads 100% as of August. Mark Williams-Cook's read is that we hit 100% shortly.
Apple shipped a Siri that answers web questions, to 2.5 billion daily requests
At Apple's September 9 event, Lilian Rincon, VP of product for AI, put a slide up saying Siri handles over 2.5 billion requests every single day, then introduced Siri AI: personal context, onscreen awareness, and broad world knowledge including answering questions from the web on virtually any topic. It is in beta in English now, with French, Japanese, Korean, Portuguese and Spanish in October. Footnote worth reading: Siri AI will not be available in the EU or China.
A leaked Google Geostore file exposes 72 local ranking signals, and the headline is that your listing is not the unit of ranking
Search Engine Land published an analysis of recovered Google binary data revealing Geostore, the internal system Google uses to represent geographic entities. Inside: 72 ranking signals under a system called Oyster Rank, 793 data source providers, and 446 local search intent types. The signals cluster into user engagement (reviews, listing impressions, direction requests), web signals (query volume, website clicks, Wikipedia), and entity characteristics (chain membership, prominence, landmarks). Notably, 25 of the 72 are marked deprecated. The author's conclusion is that the signal list matters less than Google's ability to assemble one canonical entity from many sources and bind it to the Knowledge Graph.
Scammers are uploading AI-generated photos with fake phone numbers onto Business Profiles, and Google is slow to remove them
Naomi Stevens documented a case on a roofing company's Google Business Profile where a contributor uploaded AI-generated images containing a fraudulent phone number, so customers viewing the photos call the scammer instead of the business. Sam Sarsten traced one contributor operating as "Jack Edward" with 150-plus photo contributions across unrelated US businesses, the same fake number recurring. Google's support response has been unhelpful and removal is slow. The interim workaround is to report the images and then bury them by uploading legitimate photos.
Google put AI search data inside Merchant Center: intent, terms and missing attributes
Google's AI Performance Insights in Merchant Center, launched in July and expanded regionally earlier this month, picked up three new reports spotted on September 9. AI Search Intent shows how your products match customer AI searches. AI Search Terms recommends popular terms to add to product descriptions for AI visibility. AI Attributes flags missing product attributes that are suppressing you in AI-driven results.
Jessica Bowman's argument: AI visibility is an operations problem wearing an SEO costume
In a Search Engine Land interview published September 9, Jessica Bowman argued the SEO team cannot take AI visibility to the finish line. Her example: a retailer she could not get ChatGPT to recommend, because when asked directly the model discouraged users based on complaints about product quality, returns and refunds. She notes AI systems pull from customer complaints, software release notes, earnings calls, vendor case studies and employee reviews, and that teams need to accept zero-click success, measuring mentions and narrative strength rather than clicks.
Amazon plugged its DSP into ChatGPT, and ChatGPT Ads is at a $1 billion annualized run rate
Amazon and OpenAI announced a pilot letting advertisers buy ChatGPT inventory through Amazon DSP, US only to start, with Delta Vacations among the first testers. Ads run as text and image units beneath ChatGPT responses with sponsored labels, priced on CPC or CPM, with product feed ads auto-generating assets from brand catalogs. OpenAI controls delivery through its own systems, and advertisers get aggregated impression, click and cost data. ChatGPT Ads has reached a $1 billion annualized revenue run rate with tens of thousands of advertisers.
OpenAI is refusing ads from products that compete with OpenAI
The Information reported, picked up September 10, that OpenAI has told advertising partners it will no longer approve campaigns for standalone image and voice generation products. Video generation tools remain permitted. Adobe is directly affected: it was in the initial ad pilot promoting Acrobat Studio and its Firefly image generator. Adobe's senior director for Americas media, Doug Wyatt, confirmed campaigns for standalone image and voice generation products would no longer be approved. The policy was communicated privately to partners, not announced.
Google rewired its measurement stack around first-party data and geo experiments
On September 10 Google expanded Data Manager beyond Google Ads into Analytics and Display & Video 360, adopted IAB Tech Lab's ECAPI standard to make the Data Manager API universal, and introduced a Data Strength Uplift Metric in Google Ads that quantifies the value your first-party data setup is generating. Meridian, Google's open-source marketing mix modeling platform, gained agentic capabilities that audit data quality and troubleshoot model errors, the ability to include brand signals such as Branded Google Query Volume, and a global launch of Meridian GeoX for causal geo-experiments.
Boots replaced a four-year-old platform with an imperative, and it tested as the boldest of five
Boots launched "Give it some Boots" on September 9, retiring "With you, for life" from 2022. WPP presented five concepts and this was both the boldest and the best performing in research. The strategic diagnosis is the interesting part: research found Boots had strong trust scores and strong brand power but lacked human connection, and that customers were confused by conflicting beauty and wellness advice. The hero film features women dealing with acne, hair loss, beauty confusion and menopause, scored by indie artist Gretel with choreography by Holly Blakey. Creator content and personalized offers go to 17 million Advantage Card holders, alongside a refreshed visual identity.
KFC created a global chief brand officer role that did not exist before
KFC named Amy Ellis Durini its first global chief brand officer, effective November 1, reporting to global CEO Scott Mezvinsky, with brand strategy, loyalty, retail and innovation under one seat. Durini comes from Taco Bell's international unit, where she was chief marketing and strategy officer since 2021, and spent over eight years in KFC's international business before that. The role backs a global brand transformation unveiled in June: refreshing the Colonel, redesigning stores and packaging, and a menu revamp leaning into boneless chicken, beverages and dipping sauces built on the Saucy concept. Context: KFC's international business has outperformed a US business struggling against Raising Cane's, while Yum is divesting Pizza Hut.
AG1 bought the TSA bin, because the bin is where the problem happens
AG1 put its AG1 Pro creatine launch inside airport security trays, concourse video walls and mounted glass wraps at Denver, Austin and Seattle. The copy names the exact moment: "Is your creatine travel routine raising a few eyebrows?" and "Is your DIY-packed creatine causing delays? There's a better way." The insight is the social awkwardness of carrying loose white powder through security, and the product answer is single-serve sachets.
IKEA turned its hardest-to-pronounce product names into a two-minute engagement
IKEA and Ogilvy Amsterdam launched "The Swedish Test" across the UAE, Qatar, Egypt and Oman: a TikTok voice-recognition challenge where you pronounce IKEA product names, escalating from LACK and BILLY to SKÄMTSKÅL, with gift cards for winners. The naming system exists because founder Ingvar Kamprad was dyslexic and found names easier to remember than codes. The reported result is an average of over two minutes of interaction with the brand per participant.
Tubi shipped nine spots that all open mid-scene and end at the same place
Tubi launched "Find Any Feeling For Free," its biggest global campaign, made by Mischief @ No Fixed Address with director Steve Ayson. Nine 15-second spots each open mid-story on a bizarre scenario, then pivot to the Tubi content that delivers that feeling more safely, with a consistent voiceover and a shared muted palette so the set reads as one campaign. Tubi says over 110 million people worldwide watch monthly, across the US, Canada, Mexico, Latin America, the UK and Australia, on 30-plus devices.
Google shipped its EU compliance rebuild yesterday and called it the biggest quality drop in its 29-year history
Google rolled out its Digital Markets Act redesign of European search results on September 8, and said in its own words that the changes "mark the largest reduction in quality of service at the world's most popular internet search engine in its 29-year search history." The mechanic matters more than the quote: in travel and hospitality verticals, one comparison service now sits at the top of results with two competitors below it in less detail, and any hotel, airline or restaurant carousel that follows no longer shows live pricing or real-time availability. Google says earlier rounds of DMA compliance already cut free direct booking traffic for European businesses by 30%. The rebuild follows a 460 million euro (roughly $534 million) self-preferencing fine in July 2026. Search Engine Land | Search Engine Roundtable | Quartz
Similarweb's August numbers: ChatGPT 55.5%, Gemini 25.6%, and Claude quintupled in a year
Similarweb's latest generative AI traffic share data, published September 7, puts ChatGPT at 55.5% of AI assistant website traffic, Gemini at 25.6%, Claude at 9.3%, DeepSeek at 3.4%, Grok at 2.4%, Copilot at 1.6% and Perplexity at 0.9%. The three-month move is a reversal of the story everyone was telling in the spring: ChatGPT climbed from 52.7% to 55.5% while Gemini fell back from 27.8% to 25.6%. The bigger year-over-year story is at the edges. ChatGPT is down from 73.3% a year ago and Claude went from 1.9% to 9.3%, a nearly fivefold gain. Note the caveat: this is web traffic only, so it excludes app usage and in-product AI surfaces entirely. The Decoder | Similarweb
The citation clock is 6.81 days, and 90% of pages that will ever be cited are cited within 37
Profound tracked roughly 900 newly published marketing pages across a 60-day window and found the median time to first citation by ChatGPT or Claude was 6.81 days. The 75th percentile was 18.68 days and the 90th was 37.10 days. The practical implication the researchers draw is a diagnostic one: a page still uncited past roughly 37 days is more likely to have a technical problem, a robots.txt block or a crawl issue, than a content problem. Omnibound | Machine Relations research summary
YouTube mentions correlate with AI visibility at 0.737, more than three times backlinks
An Ahrefs analysis of 75,000 brands with Domain Rating above 40, using Spearman correlations, ranked the signals associated with AI visibility. YouTube mentions led at 0.737, followed by YouTube mention impressions at 0.717, branded web mentions at 0.660 to 0.710, branded anchor text at 0.511 to 0.628, brand search volume at 0.352 to 0.466, Domain Rating at 0.266 to 0.326 and backlinks at 0.218. Separate data in the same body of research found that ranking first in classic organic results gives only a 33.07% chance of AI citation, and that over 60% of AI citations come from pages outside the top 10. The authors are explicit that correlation is not causation, and that the likely underlying driver is brand authority as a composite signal. Machine Relations | Omnibound
Google Business Profile insights have been blank for all of September, and Google is now emailing owners to confirm user-suggested edits
Nine days into the month, Google Business Profile performance reports are showing no data at all for September 2026. Separately, Google has started sending profile owners "Does this look right to you?" emails when another user suggests an edit to their listing, rather than applying the change silently. This lands alongside a broader shift the 2026 Local Search Ranking Factors survey picked up, in which behavioral signals and profile activity gained meaningful weight. Search Engine Roundtable recap | Sterling Sky local changes log
Google confirmed a test putting exact and phrase match ads into AI Mode, and Gemini-powered ad dashboards went live
Google confirmed on September 4 that it is running a small experiment serving exact and phrase match Search ads inside AI Mode, limited to queries where it sees "explicit and direct" user intent, a phrase it has not defined. Until now the only routes into AI Mode placements were AI Max, Performance Max for Search, or broad match with Smart Bidding. In the same week, the Gemini-powered AI dashboards Google previewed in May began appearing in some Google Ads accounts. Search Engine Land | PPC Land | Search Engine Roundtable
Microsoft removes Max CPC from new campaigns on October 1, and ChatGPT Ads ships with automated bidding by default
Starting October 1, Microsoft Advertising will not offer Max CPC when creating new non-portfolio campaigns using Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value or Maximize Clicks. Existing campaigns created before that date keep the setting, and Max CPC stays available for portfolio bid strategies, which is the workaround. Target Impression Share and Enhanced CPC are unaffected. Microsoft's stated reasoning is that a Max CPC ceiling sends conflicting instructions to automated bidding. The same reporting notes ChatGPT Ads launched defaulting to automated bidding, with no manual bid cap at all. Search Engine Land | PPC Land | Search Engine Journal
Magnite ran what it says is its first agentic campaign, built from natural language prompts, with a 70% setup time cut
Magnite announced on September 8 that Amnet France used natural language prompts to direct a Magnite buyer agent, which built and activated a video campaign through ClearLine for an automotive advertiser. Magnite reported roughly a 70% reduction in setup time and a 95 video view-through rate. Worth noting what is missing: there is no control comparison, and the setup-time figure is the only operational number disclosed. The Agile Brand Guide
The IAB has a measurement framework for AI visibility, and only 16% of brands are tracking it at all
The IAB's "Measuring Visibility in the AI Era" framework organizes AI visibility metrics into four categories: presence, prominence, portrayal and persuasion. The most striking finding in the accompanying work is that only 16% of brands currently track AI visibility in any form, while more than 20 vendors are selling measurement using incompatible methodologies. A companion AI attribution framework is scheduled for November 12, 2026. Seafoam monthly brief
Deloitte found back-to-school shoppers using generative AI planned to spend $737, versus $494 for search-only shoppers
Deloitte's 2026 back-to-school research found that parents who planned to use search engines for research expected to spend an average of $494, while shoppers combining traditional search, social media and generative AI planned to spend an average of $737, roughly 49% more. The category overall was forecast at $85.4 billion, with $53.06 billion in physical retail and $32.36 billion in ecommerce. StackAdapt | Eyeota
Schweppes went into a 243-year archive instead of a design brief, and brought back a bottle, a leopard and a fountain
JKR's global redesign for Schweppes, under the platform "With Time Comes Taste," works entirely by retrieval rather than invention. The team went through more than 200 years of archival material and brought back the Skittle bottle, now adapted to a one-litre PET format, the mascot Clive the Leopard, the crystal fountain illustration from the 1851 Great Exhibition, historic serif typography and the diagonal saffron sash. Rollout began in the UK and South Africa in December 2025 and continues across Latin America, Eastern Europe and Asia through 2026. The work took first place in the Soft Drinks category at the 2026 DIELINE Awards. Creative Boom | The Dieline | DesignRush
Triple Whale renamed its category on its own, from analytics dashboard to "AI operating system for modern commerce"
Triple Whale relaunched its brand identity on September 8 positioning itself as an "AI operating system for modern commerce" rather than an ecommerce analytics platform. The supporting numbers are throughput rather than outcomes, and are worth reading with that in mind: $71 billion in tracked GMV, $25 billion in ad spend, 65,000-plus brands, 100,000-plus automations, roughly one million conversations, 85-plus countries and 2,000-plus agencies. The Agile Brand Guide
The UPS Store gave 5,000 independently owned franchises one recognizable person
The UPS Store introduced Blu, its first brand character, at the start of September: a woman in a blue UPS polo who appears through a portal to solve small business owners' problems. It is a multiyear campaign across digital, TV, social and print, aimed at small business owners and built to sell the less-known services, printing, mailboxes and packing, rather than shipping. Paramount Media Labs is producing spots pairing Blu with NFL player and small business owner Aidan Hutchinson, running across CBS and Paramount streaming with digital and social content rolling out through early November 2026. Marketing Dive | Franchising.com | LBB
Aldi repainted an entire storefront Barbie pink for a £12.99 doll
To launch its £12.99 "Aldimania Dreamland" Barbie, released September 6, Aldi repainted the blue, red and orange logo on its Barnet store in Barbie pink. The doll itself is dressed in an Aldi-logo dress and visor with a miniature trolley of own-brand groceries. The stunt is a single physical location, a single day of work and a photograph that does the distribution. Famous Campaigns
Pabst turned a $70,000 beer heist into a countdown with a deadline
After roughly 40,000 pounds of beer, valued at about $70,000 across two thefts, were taken from a California distribution center in August, Pabst Blue Ribbon posted an Instagram ultimatum giving the thieves "exactly 18 days and 44 minutes" to return the truck, no questions asked, alongside a reward offer. Inc.'s assessment was blunt: the viral reaction may be worth more than the beer. Inc. | Fortune | NBC News
Spark Noir Creative curates these from each weekday's marketing reporting. Headlines and summaries
describe the original source's findings, every story links to and credits that source, and no item
is published without a link that could be verified. New stories each weekday morning.