ChatGPT Business Hub insights: for an online business, a mention is a website reference, not your brand name
Business Hub's number behaves like a citation count. Recommendations that skip your URL stay out of it, and it comes with no denominator and no competitor view.
Contents
On November 10 OpenAI opened Businesses in ChatGPT in beta. A company can claim its profile in Business Hub, edit its details, connect a product feed and look at insights; eligible US ecommerce and online service businesses can also build a Business Agent. The part that interests me is smaller. OpenAI now reports a visibility number for its own assistant, and the help article says what a mention is: for an online business, a reference to its website, not its name. By the definitions I use, that is close to a citation count, and it leaves out every answer that names you without a link.
Everything below is read from OpenAI’s launch post, landing page and help article as of November 11, not from a dashboard.
What does ChatGPT Business Hub report in its insights?
Three aggregates over the past 28 days: mentions, website clicks and discovery topics. You see them only after OpenAI approves your claim (verified one of four ways: phone, business email, a DNS TXT record or Shopify) and a separate identity check through Persona. TechCrunch also lists profile views; OpenAI’s own pages don’t. One thing to know before claiming: per the help article, only a draft profile can be deleted, and Support won’t unpublish a published one as part of a setup or ownership request.
Discovery topics are themes in answers, not prompts. The help article says they “don’t reveal individual prompts or directly measure customer intent.” Counts appear only above thresholds OpenAI does not publish, and a blank field “doesn’t necessarily mean zero activity.”
What does a “mention” mean in ChatGPT Business Hub?
For an online business, a reference to its website. The brand name on its own is not counted. The help article, as of November 11:
Mentions. For an online business, a mention counts when ChatGPT
references the business's website. The brand name alone doesn't count.
For a local business, a mention counts when ChatGPT shows a result for
that location, even if the customer doesn't open the listing or visit
the website.In my measurement workflow a mention is the name in the answer text, and a citation is a URL you own attached as a source. By that vocabulary OpenAI’s online “mention” is close to a citation. Local businesses get a different unit, a shown location result, and the article doesn’t say what a business with both a shop and locations sees.
It also doesn’t say what “references” means mechanically: a source annotation, a link written into the answer, or both. In my panel that choice was the difference between 2 and 5 citing answers out of the same 7.
Which ChatGPT answers stay out of the Business Hub mention count?
For an online business, any answer that names or recommends it without referencing its site. On the OpenAI web search surface of my September panel (the API, not the ChatGPT app), 7 of 80 non-branded answers named Beetroot, the clipboard manager I build. One of them recommended it off a forum thread, with no Beetroot page retrieved at all. Seven answers through the API say nothing about rates. They do show the mechanism: an answer can name you, even recommend you, and add nothing to a count built on this rule.
The count also arrives with no denominator and no competitor view. Forty mentions, to pick a number, doesn’t tell you what share of relevant answers included your business, and Business Hub doesn’t say. The help article also doesn’t specify which link surfaces a website click covers or how repeat clicks are counted.
How does ChatGPT Business Hub compare with Bing’s AI Performance report?
It isn’t the first vendor report of this kind. Microsoft added an AI Performance report to Bing Webmaster Tools in February 2026, covering Copilot and Bing’s AI summaries. It counts citations per URL and shows a sample of grounding queries, the phrases the AI used to retrieve the pages it cited. Search Console has had a generative AI performance report since the summer, with impressions for AI Overviews and AI Mode together. Business Hub shows themes instead of queries, and the help article describes no per-page breakdown. None of the three, as documented, compares you with competitors.
Does claiming a ChatGPT business profile change ChatGPT’s answers?
OpenAI promises nothing. The help article says claiming “doesn’t guarantee increased visibility or give you direct control over ChatGPT’s independent recommendations,” and that “advertising doesn’t influence ChatGPT’s independent answers.” Only the second is a flat denial. “Doesn’t guarantee” is not “doesn’t change”, and the landing page says ChatGPT “may use” the information you submit to answer customer questions. A product feed doesn’t “guarantee that ChatGPT recommends your products” either.
These are OpenAI’s statements. Checking them from outside would take runs in the app, not the API, before and after a claim, with every model release in between marked, and even then I’d treat a before/after difference as something to investigate, since the model moves under you anyway.
How should you read ChatGPT Business Hub numbers?
My read: as a count of website references with half a definition, not as a visibility score. The written half is still the useful part, because it tells you what the number leaves out.
For mentions, keep your own panel with the name in the text, inclusion and citation as separate fields. Business Hub adds the one thing a panel can’t, a count from real conversations in the app. Keep it as its own series next to the panel’s citation column and don’t expect the two to agree: one is a rate over prompts you chose, run through the API, the other a count over conversations you can’t see.
Website clicks is the number you can put next to something of your own. ChatGPT appends utm_source=chatgpt.com to the links it cites, and on this site in September and October the tag arrived while the referrer almost never did. Pull tagged landings for the same 28 days and report them side by side with Business Hub’s click count. Don’t expect a match: a landing counts only if your analytics loaded, and the tag was reported missing for several days in October. A large gap tells you to find out what each side counts before you report either.
01 / Discussion
No comment section here — all discussions happen on X.
02 / More articles
GPT-6 Astra and weight-space programming: what the Qwen MATH-500 result actually measured
The abstract says GPT-6 Astra reads raw weights. Three entries in the experiment table do, on networks of at most 172K weights. The 800M-parameter Qwen fix came with Python and up to fifty practice batches, and the analogous Gemma run failed. A real result, and a different one from the headline.
Gemini dropped utm_source=gemini: its iOS app traffic sits in GA4 Direct, and only a server log can count it
Three weeks of tagging changed less than it seemed. In GA4 a source-only tag drops sessions into Unassigned, an iPhone WebView click carries no referrer either way, and a raw request log is still where those clicks can be counted.
robots.txt AI training opt-out with __usage_ai-training: Google's mapping will also cover Gemini grounding
The shared token is defined as a training preference and nothing else. Google plans to read an opt-out as a Google-Extended disallow, and Google-Extended also decides grounding in Gemini Apps, so the one vendor that has said how it will read the token reads it as more than training. What each of the three mechanisms controls, how they stack, and a file per intent.