ChatGPT Now Has a Business Hub: What Verified Profiles and Business Agents Actually Change
OpenAI opened a Business Hub in ChatGPT where companies claim profiles, feed in live pricing, and build a Business Agent to answer buyer questions. Here is what is confirmed, what is still beta, and how an operator should prepare for a search surface that reads like a conversation.
TL;DR: OpenAI is letting businesses claim a verified profile inside ChatGPT, feed it live pricing and availability, and (for some U.S. ecommerce sellers) run a labeled “Business Agent” that answers buyer questions, which means the thing you optimize for is shifting from a page that ranks to a profile an assistant can cite.
The primary source here is OpenAI’s own announcement, “Introducing Businesses in ChatGPT,” plus the companion video “Introducing Businesses in ChatGPT in beta.” Both describe the same product: a Business Hub where verified companies manage their information, add product feeds, see aggregated insights, and in some cases build an agent. Everything below about what it does and who can use it comes from those two OpenAI sources. Where I am reading tea leaves, I will say so.
What did OpenAI actually ship?
Three things, by OpenAI’s own description.
First, verified business profiles. A company claims itself in Business Hub (chatgpt.com/business-hub), verifies, and then controls the details ChatGPT shows about it. OpenAI says verified businesses can “update their details and add product feeds with current pricing and availability.” That last part matters more than the profile itself. A feed with live pricing and stock is structured data, not a marketing paragraph. It is the kind of input an assistant can pull directly into an answer.
Second, insights. OpenAI says verified businesses see “aggregated insights on mentions, website clicks, and common topics in conversations where their business is cited.” Think of it as analytics for the conversational surface. Not keyword rankings, but how often you get mentioned, what people click through on, and what topics come up around you. Note the word aggregated, and note OpenAI’s explicit line: “We don’t share people’s conversations or Business Agent chats with businesses.” So you get counts and themes, not transcripts.
Third, and this is the one people will argue about, the Business Agent. OpenAI says “eligible ecommerce and online service businesses in the U.S.” can build an agent that answers customer questions, compares products, and helps people find what fits, using information the business provides. Crucially, OpenAI describes this as “a clearly labeled conversation, separate from ChatGPT’s independent answers.” So there are two lanes: ChatGPT’s own answer, and a business-controlled agent chat that is marked as such.
Who can use it, and what is still beta?
This is where I want to be careful, because the details are first-party and the product is new.
OpenAI labels the whole thing beta. The Business Agent is gated: “eligible ecommerce and online service businesses in the U.S.” Profiles and the Hub are broader. OpenAI says “eligible businesses located wherever ChatGPT is available can get started by claiming their business in Business Hub.” So claiming and managing a profile is wide, building an agent is narrow and U.S.-only for now.
What OpenAI does not give in these sources: pricing, rate limits, how verification actually works, what counts as “eligible,” review timelines, or how feeds get validated against reality. None of that is stated, so I am not going to invent a number or a rule. If you read a specific price or limit somewhere, check it is OpenAI saying it and not someone guessing. Treat eligibility and mechanics as “announced, details pending” until the docs fill in.
One thing I will flag as judgment, not fact: the separation between ChatGPT’s independent answer and the labeled Business Agent is the whole trust model. If that line blurs, the product is just paid placement wearing a conversation. OpenAI clearly knows this, which is why the “separate from ChatGPT’s independent answers” language is front and center. Whether users actually perceive the separation is an open question nobody can answer yet.
How is this different from SEO, and does it replace it?
It does not replace search. It adds a surface. But the mental model shifts.
Classic SEO optimizes a page a crawler reads and a human clicks. This optimizes a structured profile and feed an assistant reads and cites, sometimes without a click at all. The insights OpenAI describes make that explicit: you get mentions and website clicks as separate metrics. A mention with no click is a new kind of impression. You showed up in the reasoning, you just did not get the visit.
For anyone who builds and ranks sites, this is the part to internalize: the asset you maintain changes shape. It is less “a page that argues why you are good” and more “a current, accurate, machine-readable statement of what you sell, for how much, and whether it is in stock.” I run Lucky Domains, which does white-hat SEO and website builds, and the practical read on my side is that structured accuracy now competes with prose for attention. The feed that is correct beats the landing page that is clever, because an assistant can trust the feed and cite it.
The catch: you are now optimizing inside someone else’s walled surface with its own rules. Google at least shows you a page. Here, a lot of the value accrues to OpenAI’s answer, and your reward might be a mention you cannot click. That is not a reason to skip it. It is a reason to keep your own site, your own feed, and your own first-party channel strong, so you are not renting your entire presence from one assistant.
What breaks or gets gamed?
Every new surface gets gamed, so let me name the obvious pressure points.
Feed accuracy is the first. If businesses can push pricing and availability, some will push stale or misleading data. OpenAI has not described (in these sources) how it validates feeds against reality. Expect that to be the first thing adversarial sellers probe.
The second is the labeled-agent boundary. A Business Agent is the business talking, which means it has every incentive to steer. The label is the guardrail. How prominent and persistent that label is will decide whether this reads as helpful or as an ad that learned to chat.
The third is prompt injection and manipulation of the citation layer, which is a known problem for any system that pulls third-party content into answers. OpenAI does not address it in these two sources, so I am flagging it as an open risk, not a reported flaw.
Practitioner’s take: claim your profile now if you are eligible, because claiming is low-cost and verification is likely first-come in spirit even if OpenAI has not said so. Then do the unglamorous work: build or clean a product feed with real, current pricing and stock, because that structured feed is what an assistant can actually cite, and accuracy is your moat when competitors push sloppy data. Watch the insights for the gap between mentions and website clicks, because that gap tells you where you are showing up in answers without earning a visit, which is a content and positioning problem you can fix. If you are offered the Business Agent, treat it as a support channel, not a sales bot, and keep the labeling honest, because the moment users feel tricked, the whole conversational-commerce bet unwinds. The catch most readers will miss: this is beta, the eligibility and mechanics are not fully documented, and you should not rebuild your funnel around a surface whose rules OpenAI can change next quarter. Keep your own site and your own feed as the source of truth, and let ChatGPT be one more place that reads from it.