Market analysis

· 30 min

Brands in AI answers, 1–18 November 2026: the assistant starts keeping score

In the first half of November the assistants began counting brands themselves: ChatGPT reports mentions and clicks, Google splits branded queries in its API. The counts are new; what they say about buying is not much.

Topic radarSelect a topic to filter the archive. The radar shows how this issue's words divide across the topics.

Signal readout

Topic mix

  • AI: 37%
  • Google & Search: 13%
  • PPC: 12%
  • Analytics: 10%
  • Brands: 28%
  • AI37%
  • Google & Search13%
  • PPC12%
  • Analytics10%
  • Brands28%

Share of the issue's words about each topic, counted section by section.

Why it matters

If you can claim a ChatGPT business profile, claim it and record the first weeks as a baseline. Read mentions next to your own analytics, never instead of them, and keep running a fixed set of buyer questions you did not choose to flatter yourself: Business Hub tells you how often you were mentioned, not against whom, nor in answer to what.

The short answer

What
A mid-month reading of brand visibility in AI assistants and search for 1–18 November 2026: ChatGPT's Business Hub and Google's branded query data, which brands assistants name and why, how many links ChatGPT shows and how little of its use is shopping, the brand's own results page, ads and agents inside answers, licensing, paid AI-visibility campaigns and what link building and digital PR still do.
Why
Because for the first time an assistant reports a brand's mentions and clicks itself, while two independent studies say ChatGPT is rarely where people decide what to buy. The new numbers and the old question arrived in the same fortnight.
Who
Brand and marketing leads, SEO, GEO and digital PR practitioners, retailers, local businesses and the publishers who carry a brand's coverage.
Where
Mostly the US, with UK data from Sistrix, Ask Maps expanding to 30 European countries, and global changes from Google, OpenAI and Microsoft.
When
1–18 November 2026, with a few items of 30–31 October that were first reported inside the window, and October's figures where a comparison needs them.
How we know
From the November weekly digests' fact-checked sources and research notes for 14–18 November: studies with stated samples (Ahrefs, SparkToro, DEJAN, Seer, performance.io and Semrush, BrightEdge, Sistrix, Adthena, VAB), platform announcements and help pages, and trade press. Where only a write-up was read (PPC Land, Search Engine Journal, Search Engine Roundtable), the text says so. Every source is listed at the end.

In plain words

Until now, if a shop wanted to know how often ChatGPT talked about it, it had to ask ChatGPT lots of questions and count. In November ChatGPT itself started telling shops how often it named them and how many people clicked. Researchers also found that ChatGPT now shows many more links, but that people rarely use it to decide what to buy; Google is still where most shopping questions go. Google keeps moving its AI summary around a brand's own search page, ads are showing up inside AI answers, sometimes for a rival when you search for a brand, and some publishers now sell articles meant to be quoted by AI.

Our October review ended on a measurement problem. A citation lasts days, one prompt can miss a brand's true share by 34 points, and the only honest reading of a brand in AI answers is a share across many questions, run again and again by somebody outside the assistant.

The first eighteen days of November changed who does the counting. On 10 November OpenAI opened Business Hub, where a verified business sees how often ChatGPT mentioned it and how many people clicked. On 16 November Google put its branded and non-branded split of Search Console data into the API. For the first time the platforms themselves are keeping score on brands.

The same fortnight made the score harder to read. ChatGPT now puts links in nearly half its chats, but SparkToro found that only 6.9% of prompts show research toward a brand to buy from. Google keeps rearranging the brand's own results page. Ads sit inside answers, sometimes for a rival on a brand's own name. And a publisher now sells articles written to be cited by AI. This analysis follows those threads by theme and ends with what to watch in the rest of November.

The first half of November in one table

The studies and platform numbers below carry most of the argument. Their samples, markets and methods differ, and several come from companies that sell AI-visibility tools or services, which is noted where it applies. Read them side by side rather than adding them up.

Who measuredPublishedSampleFinding for brands
BrightEdge31 OctHoliday travel answers, AI Overviews and ChatGPTDifferent leaders per engine: Hilton in AI Overviews, Marriott in ChatGPT
Ahrefs3 Nov10,000 US brands, 260m non-branded promptsReddit mentions correlate 0.75 with AI visibility, news 0.67
Ahrefs11 Nov5.5m US ChatGPT answersWalmart in 95,605, almost twice second-placed Target
Seer Interactive12 NovThree prompts, 90 runs eachSeer named in 58 of 90 runs, from 0 in 2025
DEJAN13 Nov52,923 queries to Gemini 4 Hydrogen, no searchMap of 239,266 brands; Apple 100, Microsoft 94.3
Peec AI (via PPC Land)16 NovChatGPT chats, 27 Aug–14 Nov, sample undisclosedChats with an external link: ~10% before October, 47.8% on 14 Nov
performance.io and Semrush16 Nov14,300+ US pharma keywords83.8% of AI Overview citations from top-10 organic domains
SparkToro17 Nov118,163 prompts, 12,321 US panel users6.9% show purchase research; 12.1% of responses name a brand
VAB (via PPC Land)17 Nov40 brands returning to TVBranded search index up 18% across 29 brands, no control group
The evidence behind this analysis. Dates are publication dates; details, caveats and links are in the sections below and in the Sources.

October's watch list, checked

Each edition keeps its own address so that readers can judge how its readings held up. The October review ended with seven things to watch in November. Eighteen days in, this is what happened to each.

What we said to watchStatus on 18 NovWhat happened
Anthropic's rule against seeding AI answer sources, from 12 NovDate passed; no follower yetThe policy's stated effective date came; in the sources we read, no other model maker wrote the same line
Meta's Commercial Terms with the crawl clause, 20 NovWent the other wayMeta paused the updated business and ad terms that were due on 20 November (reported 10 Nov)
AI Mode ads in Brazil, Mexico and the UK; separate reportingNot yetTen more countries went live on 10 Nov; Brazil, Mexico and the UK still due this quarter; AI Mode still not reported separately
ChatGPT's "best" labels and creative enhancementsNot yetNo new report on the labels or on where rewritten ads send clicks; Business Agents became a possible ad destination
AI-written citation cards in AI OverviewsNot confirmed; cards changed otherwiseSource icons moved below the text in a test (11 Nov); three unfaded cards shown by default on desktop (17 Nov, spotted)
The next Muck Rack editionNot yetNo new edition in the sources we read by 18 Nov
Penske and Chegg: appeal or refiled claimsNot yetNothing reported in the sources we read
Status as of 18 November 2026, from the sources listed at the end. "Not yet" means nothing in those sources, not proof that nothing happened.

The scorecard is mostly "not yet", and that is a finding in itself. The items we expected to move were dated platform decisions; the one that did move, Meta's terms, moved backwards. What filled the fortnight instead were things nobody had on a list: a measurement dashboard from OpenAI and a publisher selling AI citations as a product.

The assistant starts counting

On 6 November Search Engine Roundtable showed a screenshot shared on LinkedIn by Juozas Kaziukėnas: OpenAI appeared to be building a tool to "track how often your business is mentioned in ChatGPT conversations." Four days later it was announced as Businesses in ChatGPT, in beta. OpenAI's own page refused our requests, so this is via Search Engine Roundtable and PPC Land.

A business claims its profile, verifies it by phone or DNS, edits its details and product feeds, and sees in Business Hub its mentions, profile views, clicks to its site and "discovery topics", the themes of conversations in which it was cited. Prompts and conversations are not shared, and numbers appear only above thresholds OpenAI does not disclose. US ecommerce and online-service businesses can also publish a Business Agent, a branded assistant inside ChatGPT, and "sponsor their Business Agent and use it as the destination for ChatGPT Ads".

The first dashboards that reached the press show how wide the range is.

DashboardPeriodMentionsClicksClicks per mention
OpenAI's example business7 days15,3565,432about 35%
BuildUp Bookings (live)28 days69024about 3.5%
Zianvo (live)28 days281about 3.6%
From PPC Land's write-up of 11 November 2026. The two live dashboards were labelled "Data through Nov 9 (UTC)". The share of clicks is PPC Land's arithmetic for the first two rows and ours for the third.

Clicks don't confirm website visits or conversions.

Business Hub tooltip, via PPC Land, 11 November 2026

Google moved on the same front from the search side. On 16 November it brought three Search Console capabilities to the Search Analytics API, among them a brandedClassification field that splits branded from non-branded performance. Google says "this data is only available for eligible properties"; other properties get no rows. It also adds webMultimodal, performance from Google Lens, Circle to Search and image uploads.

Third-party tracking, meanwhile, narrowed. PPC Land reported on 29 October that Semrush would stop tracking Microsoft Copilot in its enterprise AI-visibility product on 9 November because "Recent changes to Microsoft Copilot's access requirements prevent AI visibility platforms" from reading it. And the IAB's attribution blueprint of 12 November, citing Similarweb, found that only 16% of brands track AI visibility systematically, and that visits after an AI recommendation are credited to search 55.9% of the time and to direct 19.9%.

Our reading: first-party counts fix one of October's problems and leave the other. Business Hub's mentions are a census, not a sample, so the ±34-point error of a single prompt does not apply to them. But a mention is still not a recommendation, a click is not a visit, and the dashboard shows no competitor. A brand will now have two numbers that will not agree: the assistant's count of mentions, and an outside panel's share of the questions that matter. Both are needed, and neither is the other.

Who gets named, and why

The month's largest count of names came from Ahrefs on 11 November: across 5.5 million US ChatGPT answers, Walmart appears in 95,605, "almost twice as many as second-placed Target". At the top of the list, scale wins; that matches what Vaer AI found for retailers at the end of September.

Where the names come from was the subject of Ahrefs' study of 3 November. Louise Linehan compared 10,000 US brands' visibility across 260 million non-branded prompts in AI Overviews, AI Mode and ChatGPT with seven kinds of mention elsewhere on the web. Reddit came first, and press coverage did not beat ordinary web mentions.

Correlation of mention types with AI visibility, Spearman (Ahrefs, 10,000 US brands)

Reddit threads0.75
Domain mentions0.70
Webpage mentions0.70
Traffic of mentions0.70
News articles0.67
YouTube0.64
TikTok0.60
Ahrefs, "Which Types of Mentions Correlate the Strongest With AI Visibility?", 3 November 2026. AI Overviews, AI Mode and ChatGPT combined. Ahrefs: "These are correlations, so we can't say for certain that more mentions cause more AI visibility."

Try it on your own page

See what Voicescope finds on a page of yours

Everything above is measurable on your own work, and the check takes about a minute. Paste one address and read what comes back.

Open Voicescope

Free · no account · one at a time

The correlations are close together, so the useful reading is not that Reddit beats the press but that every kind of mention tracks visibility about equally: the brand that is talked about widely is the brand that is named. That fits Muck Rack's October finding that third parties supply most of what assistants cite, and Graphite's that what a model already believes about a brand predicts whether it names it.

The newest model gave that belief a shape. On 13 November, a day after Gemini 4 Hydrogen became the default model in AI Mode, Dan Petrovic of DEJAN asked Hydrogen about brands 52,923 times with web search switched off, so the answers come only from what the model learned, and built a map of 239,266 brands and 4.4 million links between them. Apple scores 100, Microsoft 94.3 and Google 90.9. Windows comes eleventh although the model "never named" it when asked for brands: it is known through the brands around it. For a brand, that is the parametric layer made visible: authority in the model is a matter of association, not only of being listed.

The engines still disagree on who leads. BrightEdge's holiday-travel study, written up by Search Engine Journal on 31 October, found Hilton leading hotels in AI Overviews at 28.3% share of voice while Marriott leads them in ChatGPT at 29.4%. Booking platforms appeared in 78% of AI Overviews answers to queries with words like "cheap" or "last-minute", against 17% of other answers, which BrightEdge calls "an observed association, not evidence that using these words causes a brand to appear."

And naming can change fast in one direction. Seer Interactive's Wil Reynolds re-ran three prompts that in 2025 made ChatGPT name 21 competing agencies and never Seer. Now "Seer went from 0 to 58 / 90 runs", and "Nine agencies that ChatGPT confidently recommended in 2025 didn't get named a single time across 270 tries." The 2025 baseline was a single run, as Reynolds says himself.

Listicles still work, just a LOT less.

Wil Reynolds, Seer Interactive, 12 November 2026

Taken together, the evidence points away from on-page tricks and toward being discussed. None of it is causal proof.

Two numbers from the second week pull in opposite directions, and a brand needs both. The first is about links. Peec AI's David Konitzny posted on LinkedIn a chart of the share of ChatGPT conversations with at least one clickable external link in the answer, reported by PPC Land on 16 November. Sidebar citations are not counted.

ChatGPT chats with at least one external link in the answer, % (Peec AI)

Before early Oct10.0
9 Oct41.0
14 Oct (peak)49.5
14 Nov47.8
Peec AI via PPC Land, 16 November 2026. "Before early October" is about 10%. The jump begins two days after a 7 October marker labelled "GPT-6 default". No sample size, prompts or market were given, and Peec AI sells AI-visibility software.

Linked chats also carry more links, about four instead of about two, and 63% of the linked URLs, 47% of the linked domains, were not among the chat's cited sources. So a brand can now be linked without being cited, and cited without being linked. Konitzny's own caution is the right one: "more external links don't automatically mean more clicks or referral traffic."

The second number is about what people use ChatGPT for. SparkToro, on 17 November, analysed 118,163 prompts in 41,961 ChatGPT conversations from 12,321 US panel users of the data provider Verso, from 1 January to 29 October 2026.

What US ChatGPT prompts are about, top five categories, % (SparkToro)

Images, visual design16.1
Information, entert.14.0
Everyday life, admin12.2
Writing, communication11.9
Health and wellness10.3
SparkToro, 17 November 2026. Only 6.9% of prompts showed research toward a brand to buy from, and 12.1% of responses named any brand, product or company. Search Engine Journal's write-up gives slightly different category shares from a census-weighted sub-sample; these are SparkToro's own figures.

Rand Fishkin estimates Google's purchase-journey query rate at at least 16 times ChatGPT's. Search Engine Journal adds that the coding was done by Claude, that the panel skews younger, female, lower-income and mobile, and that API use is excluded. Fishkin's conclusion is pointed.

AI visibility matters a huge amount! But NOT BECAUSE OF CHATGPT.

Rand Fishkin, SparkToro, 17 November 2026. He argues the weight sits in Google's AI features.

Against that stands Similarweb's estimate, restated by PPC Land on 17 November in its report on the company's results, that brands recommended by ChatGPT were 2.5 times as likely to get a visit within seven days. That is a modelled panel estimate, and Similarweb sells the data. The two do not contradict each other: a rare purchase conversation can still matter a great deal to the brand it names.

The paid side looks thinner still: a working paper by Maximilian Kaiser and Christian Schulze, written up by PPC Land on 2 November and not peer-reviewed, found that 73 of 87 advertisers who bought both ChatGPT and Google Search ads recorded no paid ChatGPT sale at all.

Our reading: ChatGPT is where a brand is increasingly linked and talked about, and Business Hub will show it. It is not yet where most buying decisions are made, on SparkToro's data. Google's AI surfaces are, which is why the next section matters more for most brands than the first.

The brand's own results page keeps moving

In October SE Ranking found AI Overviews on 82.5% of US brand searches. In November Google kept moving where that summary sits. On 2 November Search Engine Roundtable showed brand searches on desktop where the AI Overview sits in the right-hand column, in place of the knowledge panel, and the brand's own site leads the main column; the examples were Apple, Zillow, Coca-Cola, Google and Search Engine Roundtable itself. Barry Schwartz read it as Google "essentially replacing brand knowledge panels with AI Overviews". Google has not commented.

For a brand that is a better place and a riskier text. PPC Land found the first error on 3 November: Zillow's AI text gives a founding year of 2006 while the fact row in the same panel says 2004.

The citation cards moved too. On 11 November Brodie Clark saw the icons of cited sites moved below the AI Overview text in a test. On 17 November he reported that desktop AI Overviews again show citation cards on the right, fully visible without clicking into the overview, with three shown by default and a "Show All" control described as being tested; Sachin Patel called it a "sticky AI Overview." Search Engine Roundtable cites no Google confirmation for either.

Underneath the layout, the summary keeps spreading. Sistrix found AI Overviews on more than 80% of the million keywords in its UK Visibility Index on 11 November, with the newest growth on ecommerce queries: Go Outdoors had them on under 2% of its keywords until autumn 2025 and on 87.4% on 9 November 2026. The model behind them changed as well: Robby Stein said on 12 November that Hydrogen is also "being used for your hardest questions in AI Overviews", and Google said it connects people "to a broader range of helpful links than before", with no numbers. By 16 November the AI Mode header no longer named its model at all; Google's Rajan Patel called that "Intended".

Which brands those summaries cite was measured in one sector. performance.io, using Semrush Position Tracking, studied more than 14,300 US desktop keywords in five therapeutic areas, published on 16 November. 83.8% of AI Overview citations came from domains in the organic top ten, and the No. 1 organic domain was cited 68.8% of the time when an AI Overview appeared. Pharma brands' own sites held 9.2% of No. 1 rankings but received 4.2% of AI Overview citations. The study matched citations by domain, did not measure clicks, and shows associations only.

Pharma brand sites: rankings against AI Overview citations, % (performance.io)

Share of No. 1 ranks9.2
Share of AIO citations4.2
performance.io and Semrush, 16 November 2026: 14,300+ US desktop keywords, 689,100+ snapshots, AI Overview analysis from September 2024. 83.8% of AI Overview citations came from top-10 organic domains.

The pharma result is the October finding about homepages seen from the other side. Ranking earns a place in the pool the summary draws from, but the summary prefers someone else's explanation of the brand's product: a health publisher, a reference site, a government page. A brand that ranks first can still be described in the summary by a page it does not own.

The brand query itself is also a measurement of other work. VAB's study, reported by PPC Land on 17 November, followed 40 brands that returned to TV after at least a year without measured TV spend: the Google Trends branded-search index rose 18% across 29 of them, from 60.6 to 71.8. There is no control group, and the returning brands spent 2.1 times their earlier budgets, so the lift cannot be put down to TV alone. What it does show is the direction of travel: people who hear of a brand elsewhere search for it, and that search now opens with Google's summary.

Ads inside the answer, and a rival on your name

The ad side of the answer kept growing, and the studies of the fortnight say more about how brands appear in it than any announcement. On 10 November ads in AI Mode went live in ten more countries, with Brazil, Mexico and the UK still due this quarter, and multiturn ads, which "show up on responses to follow-up user queries", went live in the US. On 12 November Glenn Gabe captured a single answer with a sponsored pack of four products from local retailers plus sponsored stores; Juozas Kaziukėnas counted "Four different sponsored placements in a single response."

Seer watched 161 real AI search sessions from March to September, published on 12 November. In five of six categories a branded prompt returned a sponsored card for a different brand, and seven of twelve people who were shown an ad said there had been none. That is the scenario October warned about: a brand's own name buying a rival's placement, inside an answer the reader does not register as advertising.

Where ChatGPT ads run was counted by Adthena, which sells ChatGPT ad tracking, in a report written up by Search Engine Journal on 5 November: 204,489 placements across five categories, with 83% of beauty ads on research or discovery prompts and under half in apparel and consumer electronics. And the largest field test of the fortnight measured what AI Mode does to ad clicks. A preregistered experiment with about 1,100 US users, reported by PPC Land on 17 November from an unreviewed arXiv preprint, found that the share of users clicking an ad fell 42.7 points when searches were sent to AI Mode instead of ordinary Search. PPC Land notes that the paper does not address the shopping ads Google added to AI Mode in February.

PlatformWhat changed for brandsStatusDate
Google AI ModeAds in ten more countries; multiturn ads in the USLive10 Nov
Google AI ModeFour sponsored placements in one answerSighting12 Nov
ChatGPTBusiness Agent can be the destination for ChatGPT AdsBeta, US10 Nov
ChatGPT AdsIntelligent bidding, Target CPA and CPC; Max Bid and Max CPC retire 1 Jan 2027OpenAI email12 Nov
ChatGPT AdsAds during image generationReported10–12 Nov
BingSearch and shopping ads at the top of AI answersSighting9 Nov
Google AdsAI-generated summaries in search ads; "Found on site" labelTests9–12 Nov
Performance MaxBrand guidelines take up to three reference images for AI-made imagesSpotted17 Nov
Sources: Search Engine Journal, Search Engine Roundtable, PPC Land and Seer Interactive. A sighting is one or a few screenshots, not a confirmed launch.

Two changes put the brand's own voice inside the ad system. A Business Agent, the branded assistant a US business can publish in ChatGPT, can be sponsored and used as the destination of an ad, so the ad leads not to the brand's site but to the brand's assistant inside ChatGPT. And Performance Max brand guidelines now let advertisers upload up to three reference images as a style guide to steer how Google creates and enhances ad images. Both answer October's worry about rewritten ads in the same way: the platform will generate; the brand may steer.

Our reading: for a brand, the paid answer is now partly a matter of defence. Watch your own brand name in AI Mode and ChatGPT the way you watch it in search ads, because the rival's sponsored card can sit there unnoticed by the reader and, on Seer's evidence, unnoticed by you.

When the assistant calls, books and pays

Some answers now act for the user, and the brand on the other end is a business that gets called or ordered from. On 16 November Google said Ask Maps can find beauty, pet care, auto and home repair appointments online "and call local businesses for you", in the US first, and that Ask Maps is expanding to 30 European countries, which takes it past 200 countries and territories. Search Engine Journal points to a Business Profile setting, "Bookings and inquiries from customers", that controls Google's automated calls; Google's post does not say whether Ask Maps respects it. Earlier in the month AI Mode added direct booking links for home services in the US with Angi, Thumbtack, Housecall Pro and Jobber, and brought direct checkout to Canada with lululemon, Shoppers Drug Mart and Simons, the merchant staying the seller of record.

OpenAI launched ChatGPT Wallet on 13 November, reported by PPC Land the next day: always-on agents can pay merchants with card-network tokens, with biometric approval for each purchase, for US ChatGPT Pro users. Its Commerce Partner Playbook asks merchants to recognise requests signed with Web Bot Auth, to check that risk tools do not decline agent orders, and to keep product feeds current through Business Hub. "websites need to ensure agents are not inadvertently blocked," OpenAI writes. The same Business Hub that counts a brand's mentions now carries the feed that agents buy from.

Not every brand wants that. United Airlines changed its terms on 11 November: "No Automated Agent shall be permitted to access United's site or app without prior written permission from United", and bookings made by agents may be cancelled. For agentic commerce the brand's choice is no longer only whether to be visible but whether to be transactable, and by whom.

Who is paid to be the source

Licensing moved from lawsuits to collectives and line items. On 12 November the ILC, the collective formerly called RSL, launched with more than 3,000 media brands and one blanket licence for AI companies, priced by use, with Yahoo Scout as its first paying licensee and no rates disclosed. People Inc, which lost 26% of its core sessions year on year, reported licensing and other revenue of $52.7 million, up 49%. And Ahrefs counted on 3 November how often ChatGPT cites OpenAI's publisher partners: "Ten partner domains are among the 50 sites ChatGPT cites most often in our sample, and 47 are in the top 500," led by vogue.com with 42,108 citations and caranddriver.com with 41,603, "mostly review and lifestyle sites, not news sites".

For brands that last detail is the important one. Review and lifestyle titles are where products are compared, and the ones with a deal are among the sources ChatGPT reads most. A brand's earned coverage in those titles is coverage in the assistant's favourite reading.

The next step was to sell that reading directly. Digiday reported on 16 November that Bustle Digital Group sells a GEO offering, internally called ECHO, which uses Profound analytics and BDG's own agents to brief editors, who then write articles meant to be cited by ChatGPT, Gemini and AI Overviews, with Reddit paid media to promote them. At least three brands have paid. In BDG's case studies, built on Profound data, Innisfree's visibility across 200 tracked prompts rose 30% and Bustle's citation rank went from #709 to #12; Pampers went from #2 to #1 in AI visibility for fragrance-free diaper questions.

There isn't a defined playbook for how publishers should be approaching this.

Graham Howes, Bustle Digital Group, to Digiday, 16 November 2026. He adds: "The advertiser has no hand in the actual creative development at all."

This is the line October's manipulation section drew, approached from the legitimate side. BDG's articles are published under its own mastheads, and the advertiser does not write them; Anthropic's new usage policy is aimed at content that "misrepresents its origin, authorship, or independence". Whether a reader of an AI answer can tell a paid article from an independent one is a question no platform has answered yet. The counter-model appeared the same week: Reddit tests AI-written review pages for about 10,000 products built from its threads, and says "Businesses can't pay to have products featured".

And scale still decides who gets away with what. Google has indexed about six million AI-written pages under facebook.com/fb-answers/, which Glenn Gabe called "another example of massive 'Scaled content abuse'"; Google had not commented when Search Engine Roundtable reported it on 5 November. Brands that compete with such pages for the same questions should watch whether the policy is applied to a site of Facebook's size.

None of the three weekly digests found a link-building story worth a section of its own, and we will not invent one. What the fortnight did offer is evidence on what earns a brand its place in answers, and it points the same way as October.

Earned beats self-published. Seer's self-ranked listicles stopped working in ChatGPT, while its talks and social posts brought about 11% of its sessions and 54% of its newsletter signups; AI assistants brought 0.3% of sessions and 0.2% of signups over two years. Mentions of every kind correlate with AI visibility at similar strength, so a digital PR programme should be judged by how widely a brand is discussed, not by one type of placement. Rankings still feed Google's surfaces: 83.8% of pharma AI Overview citations came from top-ten domains, which is October's Ahrefs finding in another sector. Link building that helps the right pages rank still pays on Google's AI surfaces; links built for AI visibility alone have no new evidence behind them.

And a new channel is for sale. BDG's ECHO is, functionally, digital PR priced by AI citations. Its results are reported by the seller on the seller's analytics partner's data. Before buying, ask for the prompt set, the engines, the period and what happened after the campaign ended; October's half-life studies suggest that last question matters most.

What it adds up to

This section is our reading, not a measurement.

  • The platforms now keep score. Business Hub counts a brand's ChatGPT mentions and clicks; the Search Analytics API splits branded queries. These are census numbers with no competitors in them, so they complement outside measurement rather than replace it.
  • Being linked is not being bought. ChatGPT puts links in nearly half its chats, but only 6.9% of prompts show purchase research and 12.1% of responses name a brand. For most brands the buying decision is still made on Google's surfaces.
  • Naming follows discussion. Every kind of mention tracks AI visibility about equally, the biggest retailers are named most, a model knows brands by association, and self-ranked listicles have lost much of their power.
  • The brand's results page is Google's to rearrange. The summary moved to the right column, the cards came back unfaded, and the model behind it changed and then stopped being named. A brand ranking first can still be explained by somebody else's page.
  • The answer is an ad slot that readers do not see. Four sponsored placements in one AI Mode answer, a rival's card on a branded prompt in five of six categories, and most people unaware they saw an ad.
  • AI citations are now a product. Collective licences for publishers, partner deals that put review titles at the top of ChatGPT's reading, and a publisher selling articles designed to be cited. Coverage has a price; independence is the part that has to be shown.
  • Agents make brands transactable. Ask Maps calls local businesses, ChatGPT Wallet pays merchants, AI Mode books and checks out. Feeds and permissions are now part of visibility.

What to watch in the rest of November

  • The November 2026 core update. It began on 10 November; when Google's dashboard was read on 18 November it had no completion entry, so it was still rolling out. Google said it may take up to two weeks, and Search Engine Roundtable on 16 November expected it to finish about a week later. Read brand-term traffic changes against it before blaming AI Overviews.
  • Business Hub's thresholds and reach: whether OpenAI publishes the threshold below which mentions are hidden, and whether profiles open beyond the US and the waitlist.
  • Branded data in the API: which properties Google counts as eligible, and whether branded classification reaches the AI report.
  • Citation cards in AI Overviews: whether three unfaded cards and the right-column brand panel become the default, and whether Google confirms either.
  • AI Mode ads in Brazil, Mexico and the UK, still due this quarter, and any separate reporting for AI Mode.
  • Paid AI-visibility campaigns: whether other publishers follow BDG, and whether any platform says how it treats articles written to be cited.
  • Meta's paused terms: a new date for the crawl clause, if one is set.

To read what three assistants say about your brand across the buyer questions rather than one, and compare it with your own page, Brandscope runs them in one reading.

Questions and answers

What does ChatGPT's Business Hub show a brand?

For a business that claims and verifies its profile: mentions in ChatGPT, profile views, clicks to its site and the topics of conversations in which it was cited. It does not share prompts, conversations or competitors, numbers appear only above undisclosed thresholds, and its own tooltip says clicks do not confirm visits or conversions.

How often do people use ChatGPT to research purchases?

Rarely, on SparkToro's data: 6.9% of 118,163 US prompts from 1 January to 29 October 2026 showed research toward a brand to buy from, and 12.1% of responses named any brand, product or company. Rand Fishkin estimates Google's purchase-journey query rate at 16 times ChatGPT's or more.

Which brands does ChatGPT mention most?

In Ahrefs' count of 5.5 million US ChatGPT answers, Walmart appears in 95,605, almost twice as many as second-placed Target. Ahrefs also found that mentions on Reddit, web pages and in the news all correlate with AI visibility at similar strength, Reddit highest at 0.75.

Can Google Search Console separate branded and non-branded queries?

Yes, for eligible properties. On 16 November 2026 Google added a brandedClassification field to the Search Analytics API, bringing the branded split from the Search Console interface to custom reporting. Properties that are not eligible get no rows.

Can a publisher sell a brand visibility in AI answers?

Bustle Digital Group does: its ECHO offering briefs editors with Profound data to write articles meant to be cited by ChatGPT, Gemini and AI Overviews, promoted with Reddit ads. BDG reports gains for Innisfree and Pampers from its own case studies; the advertiser does not write the articles.

Sources

Every claim above is dated and linked. Where a number comes from a measurement rather than an announcement, whoever measured it is named; where we read a write-up rather than the original, the write-up is the source listed.

The product

Watch it on every page, every day

One reading tells you where a page stands today. The product asks the same questions of the same assistants continuously, so a change is something you are told about rather than something you go looking for.

Request an invitation

Invite-only while we keep the readings honest

On this topic

What do models believe about our brand?

Every day Competitor comparison, answer by answer. How the product works

Author

Olexandrauthor

15+ years in marketing and SEO. Dozens of products, mostly the US market. Research and measurement.

LinkedIn

Pass it on

Read in your language

Opens a browser translation of this English article. The original source stays in English.

All issues