ChatGPT tests verified business agents for shopping

By James Calloway

OpenAI is launching a beta that lets eligible U.S. e-commerce and online-service businesses create “Business Agents” that represent a brand and help with comparison shopping.

The first access is limited to eligible businesses in the United States. It launches as part of a broader suite of business-focused ChatGPT capabilities announced in beta.

OpenAI framed the release in its announcement: “Today’s announcement is an early step in bringing businesses more deeply into ChatGPT. We’re starting with the foundations: accurate business information, useful insights and conversations that help people make informed decisions.”

How does a Business Agent enter the conversation?

A Business Agent becomes available when a business that created one appears in a ChatGPT response. From that result, a user can open a separate chat, ask what the business offers, and request follow-ups such as differences between two products.

The interface then labels the exchange and separates it from the main ChatGPT conversation. Those two cues establish provenance: users can see that the respondent represents a business rather than speaking as ChatGPT’s general assistant.

That distinction matters when a shopping path contains several businesses. Rather than blending every answer into one voice, the pattern gives each company a clearly bounded exchange. The announcement describes no further visual treatment beyond the label and separate thread, so those elements carry the visible distinction.

Businesses supply each agent’s instructions and business context. Those controls let a company define how its agent answers, while the user’s questions remain within that business-specific conversation. It is a more constrained pattern than asking one general model to compare every option.

What does verification cover?

Companies can claim their organization and get it verified through Business Hub. The process applies to the organization behind the agent, giving the interface a stated ownership and verification path rather than an unlabeled business response.

At a press briefing, Benji Shomair, OpenAI’s vice president of monetization, said the company was not sharing specifics about verification because methods differ among businesses. That leaves “verified” as a bounded promise in the beta announcement: it confirms a claim and verification path but withholds the criteria users would need to assess it.

What can businesses measure?

Business Hub also gives companies insights into their mentions in ChatGPT responses and their top “discovery topics,” where the business appears. The announcement gives those as examples rather than defining the full reporting set, so Business Hub’s scope remains unspecified.

The current workflow links three interfaces: ChatGPT supplies the discovery surface, the Business Agent holds the branded conversation, and Business Hub reports visibility. That arrangement gives each stage a distinct job. Users need the label to preserve source identity, while businesses get a measurable view of where they appear.

OpenAI added: “As things evolve, we’ll explore more ways for businesses to connect with customers — directly or through agents and in ChatGPT and beyond. Our goal is to make ChatGPT more useful for people making decisions, and give businesses new ways to participate and grow.”

The next phase will test whether those labeled, business-controlled conversations remain useful as OpenAI adds more direct connections between companies and customers, and whether shoppers can still identify who is answering when more agents appear.