A verified business layer inside ChatGPT
OpenAI is adding a formal business layer to ChatGPT. Its new Businesses in ChatGPT beta combines Business Hub, where organisations can claim and maintain a verified profile, with Business Agents that can answer customer questions in a separate branded conversation. The launch turns business information from a largely external input into something eligible organisations can manage directly.
Business Hub lets a verified company update key details and submit self-service product feeds containing pricing and availability. OpenAI says eligible businesses in markets where ChatGPT operates can claim a profile, while existing Enterprise customers and advertisers can work through their account representative. Shopify merchants can connect through Business Hub or Shopify’s administration interface, building on the Shopify Catalog integration already used in ChatGPT.
How Business Agents change the customer journey
The practical change is control over the source material ChatGPT can use when people compare products or services. A maintained profile does not guarantee favourable placement, but it gives a company a structured way to correct stale details and expose current catalogue data. For ecommerce teams, the operational task will be keeping feed inventory, prices and policies aligned with the company’s own site.
Eligible US ecommerce and online-service businesses can also build, test and launch a Business Agent. When a company appears in a ChatGPT response, a user can open its profile and choose to start a separate conversation with that agent. OpenAI says the agent chat is clearly labelled and remains distinct from ChatGPT’s independent answer, an important separation for user trust.
Discovery data without conversation access
That design creates a new hand-off point between general AI discovery and a vendor-controlled interaction. A shopper could move from a neutral comparison into questions about product differences, suitability or availability without leaving ChatGPT. Businesses can sponsor the agent and use it as a destination for ChatGPT Ads, but sponsored placements are labelled and separated from model responses.
Business Hub also reports aggregated engagement information, including mentions, website clicks and broad discovery topics in conversations where the company’s website is cited. OpenAI states that it does not provide businesses with individual user conversations or Business Agent chats. Insights appear only after interaction thresholds are met, limiting the risk that low-volume reports identify a particular person.
What organisations should evaluate
For marketing and search teams, the discovery topics may become a useful signal about how a brand is represented in AI-assisted research. They should not be treated as a substitute for analytics, customer research or search data. The useful workflow is to compare the topics with verified product information, improve unclear source pages and measure whether corrected information produces better-qualified visits over time.
Organisations considering the beta should establish ownership before claiming a profile. Product, commerce, legal, privacy and support teams need clear responsibility for feed freshness, agent instructions and escalation paths. A conversational agent that answers confidently from an outdated catalogue could create more friction than a simple link, particularly where price, eligibility or regulated claims change frequently.
Teams should also test the boundary between a business-controlled agent and ChatGPT’s general responses. Labelling helps, but customers may still assume continuity between the two experiences. Published policies, traceable product data and a reliable route to a human remain necessary. OpenAI says it plans broader markets and categories, so the beta is best viewed as an early operating model rather than a finished universal business directory.
The launch also changes the optimisation conversation around AI discovery. A company cannot meaningfully improve its representation through slogans alone; it needs machine-readable facts that agree with public product pages, policy documents and live commerce data. Conflicting feeds could undermine both customer confidence and the usefulness of OpenAI’s aggregated insights, so content governance becomes an engineering responsibility as well as a marketing task.
Privacy claims deserve continuing scrutiny as the service grows. Thresholded aggregate reporting reduces the chance of exposing individuals, yet businesses will want to understand how discovery topics are formed, retained and filtered. Regulated organisations should document what their agent records, where those records live and which internal teams can access them. Customer-facing disclosures should explain when a person has entered a vendor-controlled conversation.
Australian organisations can claim eligible profiles where ChatGPT is available, but Business Agents initially target eligible businesses in the United States. That distinction should be clear in launch planning. Global brands may prepare verified data now while waiting for local agent availability. They should avoid building critical support journeys around a beta whose categories, geographic coverage, analytics thresholds and commercial terms may change.
Measurement should separate profile impressions, independent citations, agent conversations and paid traffic. Combining them into one success number would blur very different customer intentions. Teams should establish a baseline before claiming the profile, use tagged destinations and monitor whether agent interactions resolve questions or merely add another step. That evidence will show whether the new surface improves discovery, service or only visibility.