Brands in AI answers, October 2026: being cited is not being chosen

Brands spent the first half of 2026 asking whether AI assistants mention them. The studies published since July answer a different question, and the answer is less comfortable: a brand can supply the page an assistant reads, appear in its citations, and still not be the brand the answer recommends. Being cited, being chosen and being advertised have become three separate outcomes.

This analysis follows that split through the year's data: who gets named, what assistants read, how long any of it lasts, what has happened to the brand's own search results, and how ads, checkout, licensing and new manipulation rules are changing what an answer is. It ends with our reading of what a brand should measure.

A short list everywhere, a long tail somewhere

Semrush's AI Visibility Index, built from 126 million US prompts between January and April across ChatGPT, Gemini, AI Mode and AI Overviews, found only 36 brands in the top 100 most-mentioned on every platform in every month: YouTube, Google, Reddit, Amazon, Walmart, Apple and Nintendo among them. Everyone else's visibility depended on the platform. The platforms also differ in how many brands they name and how many sources they cite.

Semrush's first UK benchmark, published on 13 October, sharpened the point. Any two platforms shared only 63 to 70 of their top 100 mentioned brands in the UK, and fewer than half of the most-mentioned brands were also among the most-cited domains. Tesco and Asda were both top-100 mentions on all four platforms every month; in July Tesco ranked 25th among domains cited by ChatGPT and Asda 606th. Fractl's index, reported in August, found that SEO authority tracks AI visibility for about nine brands in ten, but 471 brands with strong SEO were still underexposed in AI answers.

Cited is not chosen

The clearest evidence came from retail. Vaer AI's study for Lightspeed Commerce, published on 29 September, analysed about 500,000 shopping answers from ChatGPT, Gemini, AI Mode and AI Overviews in the US and Canada. Without web search, assistants named chains 63–70% of the time. Shown a large and a small retailer without labels, models picked the larger one 90–94% of the time, and about ten chains took 40–44% of all recommendations. Cited sources split roughly evenly between big and small stores, about 38% each, yet chains won the lead recommendation about two and a half times as often. One word in the prompt changed the result: asking for "independent" shops lifted small and local stores from about a third to nearly four-fifths of those named, while "local" and "near me" did little.

Advertising shows the same separation from the other side. SE Ranking's study of more than 50,000 commercial prompts found sponsored placements in 25.94% of ChatGPT answers, but only 3.63% of advertisers were also cited as an organic source. Paying to appear and being trusted as a source are, so far, almost independent.

What the assistants read

The sources behind those answers moved sharply this year. Ahrefs compared 2.4 million US ChatGPT answers from March with 3.8 million from September: Reddit's share of answers citing it fell from 19.1% to 2.3%, Wikipedia's from 14.1% to 2.9%, PubMed rose from 64th to 4th, WebMD fell from 18th to 79th, and Britannica and Investopedia left the top 200. Ahrefs' reading is that ChatGPT now prefers original sources over sites that summarise them. Only 14.4% of free, logged-out responses ran a web search at all, and when ChatGPT searched a specific site among the top domains, three times in four it searched a brand's own site.

Across engines the picture differs. AirOps, analysing about 3.5 billion citations from August 2025 to June 2026, found owned brand and product pages lost 10% of their citation share while creator and social sources gained 140%, with YouTube up 158% to 4.6% of all citations. For local searches Yext found 80% of citations pointing to brand websites or listings in the first quarter, down from 88.2%, and found Anthropic's Claude drawing 19.5% of its citations from reviews and social, about seven times OpenAI's 2.6%. BrightLocal's September study of about 200,000 local prompts found business websites made up 42% of citations and Google Business Profile 28.6%, and that two runs of the same prompt overlapped only 20–33%.

Our reading: there is no single source strategy for "AI". Each engine has its own diet, and the diet changes. Publishing original data, tests and specifications — the material others summarise — looks like the most durable position, but that is one engine's pattern over one period.

Visibility has a half-life

Profound followed about 883,000 pages across seven engines for a year, to 8 September. The median page lost half of its peak citation share 11 days after the peak, and 78% of page-engine pairs had a half-life of two weeks or less. How long a page stayed cited on one engine said almost nothing about another: correlations ran between 0.03 and 0.09. Pages cited by five or more engines were five and a half times as likely to keep more than half their peak share after eight weeks.

A brand audit taken once describes one engine on one day. Visibility in AI answers behaves less like a ranking and more like a weather reading.

The brand's own query is no longer its own

The one results page a brand used to control has changed. Chris Long found AI Overviews on 93% of about 100 brand-name searches he tested, usually below the brand's own listing; DemandSphere said AI Overviews on branded queries tripled in September. A summary written by Google from third-party sources now sits on the page a customer sees when they search for the brand by name.

Where the questions go is changing too. Comscore's US panel shows ChatGPT's share of AI prompt volume falling from 70% in January to 50% in June, while Gemini rose from 17% to 30% and Claude from 2% to 11%. The clicks did not follow: in BrightEdge's August data ChatGPT still sent 95.1% of AI referral traffic, and Gemini 2.4%. A brand that judges AI by its referral traffic sees the assistant losing share of questions and very little of the two gaining it. Our AI analysis covers this in detail.

The answer becomes a shop window

ChatGPT ads, tested in the US from 9 February, reached 31 European markets on 24 August and more than 60 countries by 23 September, for Free and Go users. On 6 October ChatGPT was seen testing labels on organic product results: a green "best all-rounder" and grey labels such as "best for beginners". OpenAI has not announced it, and nothing says how the labels are chosen.

Checkout is moving into the answer on Google's side. On 22 September Merchant Center told Shopify merchants that native checkout inside AI Mode and Gemini, built on Google's Universal Commerce Protocol, was on by default, with an opt-out in the Shopify admin. OpenAI went the other way in March, stepping back from Instant Checkout to let merchants use their own checkout while it focused on product discovery. For a brand, agentic commerce makes the accuracy of prices, stock and terms in machine-readable form a visibility question.

Who gets paid to be the source

Publishers are being paid case by case rather than by right. On 30 September Judge Amit Mehta dismissed Penske Media's and Chegg's federal antitrust claims over AI Overviews, writing that "an expectation is not an agreement"; we covered the ruling in the 6 October issue. Licensing pools keep growing: CNN, Axios and the Financial Times joined Google's paid AI news programme, now more than 400 publications, on 15 October; Reach joined Google's pilot on 12 October; OpenAI signed the Times of India and Indian Express on 23 September. The suits continue alongside: USA Today Co sued OpenAI on 9 October, seeking $250 million.

Our reading: if licensing shapes which sources assistants prefer, the publishers that brands rely on for third-party coverage will matter more in some engines than others. That link is plausible, not proven.

Manipulation now has rules

Attempts to steer assistants have been documented and are now named in policy. Microsoft's security team reported in February that, over 60 days, it found more than 50 hidden prompts from 31 companies in 14 industries, embedded in "Summarize with AI" links, instructing assistants to remember the company as a trusted source. Anthropic's updated usage policy, effective 12 November, bans using Claude to manipulate the sources AI systems draw answers from by seeding content that misrepresents its origin, authorship or independence. Google's documentation, rewritten at the start of October, names fabricated author profiles as deception.

Budgets are ahead of measurement. Digiday's August roundup of surveys found marketers putting about 24% of search and content budgets into AI visibility, while Semrush's survey of 481 marketers found 45% cannot properly measure it and only 9% can measure everything they consider relevant.

What it adds up to

This section is our reading, not a measurement.

• Three results, not one. Cited, recommended and advertised are separate outcomes with separate causes. Reporting one as "AI visibility" hides the other two.
• Scale wins by default. Assistants favour large brands unless the user's own wording asks otherwise. Small brands win where the question names what makes them different.
• Originality outlasts summary. ChatGPT's shift toward primary sources rewards brands that publish first-hand data, tests and specifications.
• Visibility decays in days. An 11-day half-life and near-zero agreement between engines make repeated, engine-by-engine measurement the only meaningful kind.
• The brand query is shared territory. AI Overviews on branded searches mean what third parties say about a brand now appears on the brand's own results page.

What to watch

• Whether ChatGPT's "best for" labels ship, how they are chosen, and whether brands can affect them.
• Whether the fall of Reddit and Wikipedia in ChatGPT's citations holds under GPT-6, and whether other engines follow.
• How many merchants keep Google's on-by-default checkout, and whether OpenAI returns to checkout inside the chat.
• Whether licensing deals measurably change which sources are cited.

To read what three assistants say about your brand and compare it with your own page, Brandscope runs the buyer questions in one reading; Reviewscope does the same for what they say about your reviews.

Sources

Every claim above is dated and linked. Where a number comes from a measurement rather than an announcement, whoever measured it is named.

• Semrush: 36 brands win AI visibility everywhere — PPC Land on Semrush, 30 June 2026
• Semrush 2026 AI Visibility Index: first UK benchmark — Semrush, 13 October 2026
• Fractl's AI Visibility Index — MarketScale on Fractl, August 2026
• AI retailer bias study — Vaer AI for Lightspeed Commerce, 29 September 2026
• ChatGPT ads appear on 26% of commercial prompts — Search Engine Land on SE Ranking, August 2026
• ChatGPT's most cited sites shift to primary sources — Search Engine Journal on Ahrefs, 12 October 2026
• The fastest-growing source in AI search — AirOps, 11 August 2026
• AI citation behaviour across models — Yext Research, July 2026
• Local AI visibility study — BrightLocal, 16 September 2026
• The half-life of an AI citation is 11 days — Profound, 30 September 2026
• AI Overviews on large brand names — Search Engine Roundtable, 30 September 2026
• ChatGPT loses 20 points of US AI prompt share — PPC Land on Comscore, 22 September 2026
• ChatGPT sends 95% of AI referral traffic — The Next Web on BrightEdge, 24 September 2026
• OpenAI passes 60 countries for ChatGPT ads — PPC Land, 23 September 2026
• Product labels on ChatGPT results — Search Engine Roundtable, 6 October 2026
• Google native checkout emails — Search Engine Roundtable, 22 September 2026
• What it means that the leader in agentic commerce pulled back — Forrester, March 2026
• Penske AI Overviews lawsuit dismissed — Press Gazette, 1 October 2026
• CNN, Axios and other publishers join Google's AI news programme — Bloomberg via BNN Bloomberg, 15 October 2026
• News publisher AI deals and lawsuits tracker — Press Gazette, updated 12 October 2026
• AI recommendation poisoning — Microsoft Security Blog, 10 February 2026
• Usage policy update — Anthropic, 8 October 2026
• How marketers are building the infrastructure for AI search — Digiday, 7 August 2026