Businesses in ChatGPT is live in beta as of 10 November 2026. Eligible businesses can claim a profile in Business Hub, update their details and see aggregated insights on mentions and website clicks. OpenAI’s announcement says Business Agents, which let people chat with a business inside ChatGPT, start with eligible US ecommerce and online service businesses.
Key takeaways
- Business Hub is live in beta, counting website mentions, clicks and discovery topics
- Business Agents launch for US ecommerce and online services only, and can be sponsored
- A 2026 preprint finds ChatGPT names remembered brands, even without a supporting page
- Claiming a profile shouldn’t change how often your business appears, says OpenAI
FirstMotion tracks how ChatGPT, Copilot, Google AI Overviews and AI Mode retrieve and name brands, and works with B2B software, fintech and legal tech companies. We test every new data source before we trust it, so we checked OpenAI’s launch against FirstMotion’s tracked prompts, two research preprints and what Bing and Google already offer.
What is Businesses in ChatGPT?
OpenAI pitches the beta at the 1.2 billion people it says use ChatGPT each week. It plans to expand to more markets and business categories.
- Business Hub lets a verified business update key details, add self-service product feeds and see aggregated insights
- Business Agents let people chat with a business about its products or services in a separately labelled conversation
- Enterprise customers and advertisers claim through their OpenAI representative, and Shopify merchants connect their store from Business Hub or the Shopify admin
- Sponsored placements are labelled, according to Reuters’ report
- Any eligible business where ChatGPT is available can claim a profile, though OpenAI’s policy says not every business can
- OpenAI’s product page says eligibility depends on industry, location and its verification and integrity checks, that you must own the business or be authorised to represent it, and that a Shopify store isn’t required
- Unite.AI’s summary puts a website requirement on initial claiming
- Business Hub is free for now, and OpenAI hasn’t ruled out a paid model, per Digiday’s launch report
How did the launch unfold?
| Date | What happened |
|---|---|
| 31 Jul 2026 | ChatGPT Ads Manager shows an Agent campaign type for select advertisers, spotted by Juozas Kaziukėnas |
| Aug 2026 | Kaziukėnas sees the first steps of the analytics tool, reported in Search Engine Roundtable’s first report |
| 16 Sep 2026 | OpenAI announces a Sponsored Agents test with select US advertisers in OpenAI’s September announcement |
| 6 Nov 2026 | The “Claim your business” banner appears in Kaziukėnas’s banner post |
| 9 Nov 2026 | David Konitzny of Peec AI publishes a browser code analysis of profiles, ownership states and Business Agent entry points |
| 10 Nov 2026 | OpenAI announces Businesses in ChatGPT in beta |
Source: Search Engine Roundtable, OpenAI, Kaziukėnas on LinkedIn and Konitzny’s code analysis.
Search Engine Roundtable’s original post called it a Search Console style tool, then added a live update four days later. The launch delivered all three banner promises. Agents are limited to US ecommerce and online services at first. The Ads Manager sighting and the code analysis both predate the announcement, which makes their details pre-launch observations.
What do the Business Hub insights show?
| Metric | What it counts |
|---|---|
| Mentions | Online business: ChatGPT references its website. Local business: ChatGPT shows a result for that location. Naming the brand alone doesn’t count |
| Views | Listed by OpenAI via Reuters, with no definition in the material I reviewed |
| Website clicks | Clicks on links from supported ChatGPT surfaces, which don’t equal completed visits or purchases |
| Discovery topics | Common themes in conversations where ChatGPT cites your site |
| Not included | Individual prompts or chats, and share of voice |
Source: an OpenAI spokesperson and Shomair, as Digiday reports; Reuters.
A smaller B2B brand might see little or nothing, because low-volume counts can be withheld. OpenAI declined to say how long mentions are tracked, though one ad executive with access told Digiday the chart covers the past 7 or 28 days.
I haven’t seen a populated dashboard, and Search Engine Roundtable’s live-launch report shows its own claim still pending with no data. Read the chart as a count of website references. How often ChatGPT names your brand stays hidden.
Who can use Business Agents?
- OpenAI says agents start with eligible ecommerce and online service businesses in the US, with access expanding to more types of business over time, and hasn’t defined “online service”
- Benji Shomair, OpenAI’s VP of monetization, declined to give Digiday a timeline for wider rollout
- Unite.AI says verification alone doesn’t guarantee agent eligibility, because OpenAI also weighs business category and intended use
- Agents are built from the instructions and product feed a business supplies, and businesses can’t bring their own
- Shopify merchants share product details, pricing and availability automatically through Shopify’s catalog, and preview the agent before they publish it
- Brands get conversation themes instead of transcripts
- Sponsored agents are confirmed, and Shomair told a press briefing that the business agent is the organic experience and the sponsored agent is the paid one, with the same agent behind both
- Kaziukėnas described MCP tools and lead capture forms, which I haven’t seen OpenAI confirm, and as of 31 July he hadn’t seen those ads live
Whether B2B software counts as an online service is unconfirmed.
How does the profile work inside ChatGPT answers?
David Konitzny, a GEO researcher at Peec AI, published a code analysis on 9 November, the day before launch. Peec AI sells AI visibility tracking, which makes this vendor research. It describes what the interface was prepared to display, and live behaviour may differ.
- Profiles sit on ChatGPT’s entity system and open in a sidebar when a user clicks a highlighted brand name in an answer
- A profile loads only with an entity knowledge identifier, 1 of 4 business-like categories and an active feature gate
- Profile fields include logo, colour, website, delivery and returns policies, suggested questions and a “Start chat” button
- The code shows claim, dispute and view request states, plus a verified badge that Unite.AI says now appears as a blue check
- Four entry points lead into the agent conversation, including a composer pinned ad
OpenAI says any eligible business can claim a profile, while the code analysis suggests organic loading depends on entity recognition. Unite.AI says users open a profile when the business appears in a model response, and Shomair told Digiday that claiming shouldn’t change how often a business appears. I wouldn’t assume claiming alone puts a profile in answers.
How does it compare with Bing and Google?
| ChatGPT Business Hub | Bing Webmaster Tools | Google Business Agent | |
|---|---|---|---|
| Reporting | Website mentions, views, website clicks, discovery topics | Citations, grounding queries, Intents, Topics, Citation Share, period Compare | Performance insights listed as coming soon |
| Competitor data | Share of voice isn’t part of the beta | Citation Share doesn’t expose competitor domains | Not described |
| Branded agent | Business Agents, US ecommerce and online services first | Not part of the tool | Yes, customised in Merchant Center, Shopify or Search |
| Status | Beta, free for now, launched 10 Nov 2026 | Preview since Feb 2026, new features June 2026 | US rollout, needs a US ecommerce store |
Source: OpenAI, Reuters and Digiday for ChatGPT; Bing Webmaster blog, June 2026; Google Business Agent help page.
According to Bing’s June update, Citation Share is the percentage of citations a site receives out of all citations shown for the same grounding query. Microsoft frames it as an observational metric with no competitor domains or traffic share behind it, and the tool targets publishers and site owners.
Google’s Business Agent lets shoppers chat with a brand on Search, and its help page lists the entry requirement: a US ecommerce store with a verified Merchant Center account. That makes ChatGPT’s and Google’s agents both ecommerce-first. Bing’s reporting sits outside that pattern.
What does research say about how ChatGPT decides which brands to name?
Chen and Koudas’s preprint, submitted on 10 October 2026, hasn’t been peer reviewed. Its findings are observational, drawn from 158,099 answers and 468,037 citations across two contrasting markets.
- Which pages ChatGPT’s own search retrieves explains far more of the variation in citation than page properties do, at 32% against 2.8%
- ChatGPT writes brands it knows into its searches, prefers websites it knows and names brands without any supporting page
- The authors say the findings replicate in a market of small B2B vendors
A second preprint reports that retrieval visibility isn’t citation visibility. A researcher’s blog summary adds that hostnames and entities stay steadier across runs than exact URLs, and stresses that the findings are observational associations.
My read is that claiming gives OpenAI a new input about you, though OpenAI says it shouldn’t change how often you appear. I checked the pattern against our own data.
What do my tracked ChatGPT prompts show?
I checked 3,224 non-branded ChatGPT answers tracked in Peec between 13 Sep and 12 Oct 2026. The same domain, firstmotion.com, behaves differently on each measure.
| Measure | Share of tracked non-branded answers |
|---|---|
| firstmotion.com retrieved as a source | 15.0% (483 answers) |
| FirstMotion named in the answer | 8.1% (261 answers) |
| firstmotion.com cited inline | 6.5% (209 answers) |
Source: Peec, ChatGPT UI, tracked non-branded prompts, 13 Sep to 12 Oct 2026 (3,224 answers).
Retrieval is the widest funnel, and being retrieved doesn’t guarantee a citation. ChatGPT can also name a brand without citing its site, as Chen and Koudas found. These numbers can’t show how the named and cited answers overlap.
Branded prompts distort the picture. Across all 3,434 answers FirstMotion was named in 13.6%, but prompts that put the brand name in the question produced a named result in 207 of 210. Excluding them leaves 8.1%. Those are tracked prompts, and Business Hub’s counts will differ from them.
What does ChatGPT search for alongside FirstMotion’s name?
- Peec records the sub-queries ChatGPT rewrites each prompt into
- Across 5,678 sub-queries for non-branded prompts, 44 (0.8%) named FirstMotion or ContextualJourney, though no prompt asked about the brand
- I reviewed 10 of the 44, all single occurrences, and 6 paired the name with “case study” or “clients”, for example “FirstMotion” “B2B SaaS” clients GEO case study
- I can’t tell whether ChatGPT knew the brand from memory or picked it up from a page it retrieved earlier in the same answer
That fits the preprint’s finding that ChatGPT writes brands it knows into its searches. It suggests ChatGPT looks for proof when a name comes up. Case studies and client evidence need to be easy to retrieve. Our methodology explains how we track prompts, and the wealth manager case study shows the client’s site retrieved in 9.8% of tracked chats in August, up from 1.7% in March (Peec, 125 tracked prompts, March to August 2026).
Which sources does ChatGPT use where competitors are named and FirstMotion isn’t?
Peec lists domains ChatGPT retrieved in non-branded answers where at least 2 competitors were named and FirstMotion wasn’t. The third party sources on that list are the ones to act on.
| Source | Type | Tracked answers retrieved in |
|---|---|---|
| seocompare.co.uk | Agency directory, self-described as independent | 19.8% (637) |
| goodfirms.co | Review directory | 3.3% (105) |
| g2.com | Review platform | 2.8% (91) |
Source: Peec gap report, ChatGPT UI, tracked non-branded prompts, 13 Sep to 12 Oct 2026.
seocompare.co.uk was retrieved in more answers than firstmotion.com itself, at 15.0%. Retrieval isn’t influence, but a domain pulled into a fifth of category answers deserves a look. SEOCompare’s about page discloses a commercial relationship with Rank4AI, which also appears on the gap list.
What should B2B brands do now?
Shomair told Digiday that claiming shouldn’t change how often a business appears. The profile alone won’t change who ChatGPT recommends, so the work sits upstream, in what ChatGPT retrieves and remembers about you.
- Check eligibility and try to claim. OpenAI says not every business can, and Search Engine Roundtable verified through DNS, but Jingletree’s report quotes Shomair saying OpenAI isn’t sharing specifics because methods differ by business
- Keep your information accurate, since it gives ChatGPT a clearer picture of what you offer, which Shomair says may help it surface you in more relevant conversations
- Test entity recognition by running 10 to 15 buyer prompts for your category and noting whether your brand shows as a highlighted entity or plain text
- Make your facts consistent across your website, LinkedIn, funding announcements and review profiles
- Publish case studies and client proof with clear titles, because ChatGPT searches for them by name
- List the comparison sites and directories ChatGPT retrieves where competitors appear and you don’t, then fix those listings
- Baseline by buyer role (founders, investors, marketers) and report non-branded visibility as the headline
What are the open questions?
| Question | Status |
|---|---|
| Does B2B software qualify? | Unconfirmed. OpenAI says eligibility depends on industry, location and its verification and integrity checks, and hasn’t defined “online service” |
| How is ownership verified? | Methods differ by business, per Shomair |
| How long are mentions tracked, and what are the volume thresholds? | OpenAI declined to say how long mentions are tracked and hasn’t published thresholds |
| What will Business Hub cost? | Free for now, with a paid model not ruled out |
| How does claiming interact with entity recognition? | Unknown |
| How are disputes handled? | OpenAI says it doesn’t resolve every factual dispute, and the code shows dispute states |
Source: OpenAI product page and policy, Digiday, Reuters and Konitzny’s code analysis.
There’s a risk to watch. Search Engine Roundtable’s author said he was nervous about people hijacking businesses through the claim flow, and OpenAI’s policy says it doesn’t verify every detail or resolve every factual dispute.
What I’d tell a founder this week
Check whether your business is eligible and try to claim it. OpenAI says eligibility depends on industry, location and its own checks, and it hasn’t said whether B2B software counts. Search Engine Roundtable verified through DNS, but Shomair said methods differ by business.
Don’t treat the chart as your visibility. It covers ChatGPT alone, and OpenAI counts references to your website, which means naming your brand won’t register. Keep tracking the other engines alongside it.
Ignore any number that includes your own name in the prompt. Branded prompts inflated FirstMotion’s ChatGPT visibility, and the extra points told me nothing. Set your baseline now, before the hub’s data makes it easy to fool yourself.
Find out whether ChatGPT names your brand
If you want to know whether ChatGPT names your company, and which competitors it names instead, FirstMotion’s GEO agency work starts with mapping the prompts your buyers use. Citation building and entity fixes follow.
Book a call and I’ll show you where you stand.