Publishers selling generative engine optimisation can now audit ChatGPT Business Hub as a first-party ruler, not a ranking boost. The 10 November 2026 beta gives verified businesses aggregated mentions, website clicks and discovery topics that sellers previously estimated from outside. This guide covers what those figures count, how blanks differ from zeros, and what to check before a number enters a client deck.
Updated November 11, 2026
Key Takeaways
- Business Hub launched in beta on 10 November 2026.
- Online mentions count a website reference, not the brand name.
- A blank can mean too little activity or unavailable reporting.
- A product feed does not guarantee ChatGPT recommendations.
What should you confirm before anyone quotes a figure?
Tomorrow’s Publisher framed the beta as a test for Future and the Daily Mail: advertisers can challenge a campaign report, and a count-only supplier faces a harder sell.
- Separate the products. Profile claiming is the broader offer. A Business Agent is a separately labelled conversation, limited to eligible US online shops and online service businesses with a website.
- Confirm the client can see numbers. TechCrunch reported a formal process. Benji Shomair, OpenAI’s VP of Monetization, described multiple methods that differ by business type and market. Tech-Ish listed a phone code, an email code, a DNS TXT record or a Shopify login, then Persona identity checks before any figures.
- Check eligibility. Physical stores and restaurants can currently claim only an existing US listing. Availability “depends on the business’s location and type,” and OpenAI publishes no country list.
- Read the terms before the rate card. OpenAI’s Businesses in ChatGPT terms, updated 9 November, say OpenAI owns Insight Data. They permit measurement, analysis, optimisation and reporting, including by a service provider for that business. That is not a right to resell the numbers or build a shared dataset unless OpenAI authorises it in writing.
What do ChatGPT Business Hub metrics actually count?
Tech-Ish reported mentions, website clicks and discovery topics over the past 28 days. TechCrunch also listed profile views among limited insights. Do not fuse those labels into one score.
| Signal | What it counts | What it is not |
|---|---|---|
| Mention, online business | A reference to the business website | The brand name alone |
| Discovery topics | Themes where the site came up | Individual prompts or measured intent |
| Blank figure | Too little activity, or reporting unavailable | Automatically zero |
| Product feed | Uploaded products, prices and stock | A guarantee of recommendations |
Tomorrow’s Publisher, citing OpenAI’s help page, is the source for the mention rule and the blank-versus-zero distinction. A figure appears only once enough conversations have built up.
Why will a prompt panel and the dashboard disagree?
A fixed prompt panel is an experiment the publisher designed. Hub counts follow OpenAI’s definitions and limits, so the two series should not be expected to match. Investigate the methods before calling a gap a win or a failure.
Jessica Davies reported in Digiday on 26 October that the Daily Mail had 12 to 15 tests running across the UK, US and Australia. Carly Steven reported visibility in one unnamed personal-finance test rising from about 50% to nearly 75% across target prompts. That set is not all customers, and it does not establish sales.
A hypothetical in that Tomorrow’s Publisher piece: rising from 50 to 75 of 100 monitored answers is 25 extra appearances, not 25 shoppers. Put Hub figures next to the panel and next to post-visit records, the same split as a citation-to-click audit.
What can a GEO pitch still sell?
The dashboard does not end work publishers can sell. It makes a count-only service easier to question. Farhad Divecha, group chief executive of Accuracast, told Digiday: “If their estimation data was wildly erroneous, that will make us rethink paying for those tools.” That is a condition, not proof a named supplier failed.
Sara Guaglione reported in Digiday on 1 October that Future had more than 30 clients and renewals. Ziff Davis has also reported clients. Those are signs of demand, not an audit of market size or margin.
Sell interpretation. Connect a repeatable test, the Hub report and post-visit records, against untreated content or a baseline. That is not perfect attribution, and it is not a promised position: a feed does not guarantee recommendations, claiming does not change whether a profile appears, and ChatGPT can keep using other sources.
The labelled agent and sponsored placements sit apart from independent answers. A client list does not reveal the margin.
Frequently asked questions
What does a Business Hub mention count?
For an online business, OpenAI counts a reference to the website, not the brand name alone.
Does a blank metric mean zero?
No. A blank can mean insufficient activity or unavailable reporting, rather than zero.
Does a claimed profile or product feed boost recommendations?
No. A feed does not guarantee recommendations, and claiming does not change whether a profile appears.
Can a publisher pool Business Hub numbers across clients?
OpenAI owns Insight Data. Reporting for one client is permitted; reselling the numbers or building a shared dataset needs written authorisation.
