Written by Admin Alex · Fact-Checked by M.Ali · Info Verified September 2026
We review and update this article regularly as new information becomes available.
TL;DR: OpenAI is testing a new ChatGPT ad format called Sponsored Agents, which opens a separate, clearly labeled conversation after an ad appears, letting users ask follow-up questions about sizing, features, or compatibility before clicking through to buy. It’s live with a small group of US advertisers now, and Shopify and HubSpot are the first integration partners.
An ad you can interrogate. That’s the new pitch from OpenAI, and it’s a stranger idea than it sounds at first.
Here’s how Sponsored Agents work: you see an ad inside ChatGPT, and if you have questions, you can open a separate conversation to ask them. Does this jacket run small? Will this software work with my setup? The AI answers, using information the advertiser provided, and if you’re convinced, it links you to the advertiser’s site to finish the purchase.
Why OpenAI drew a hard line here
The company is being unusually specific about where the boundary sits. Sponsored Agent conversations are separate from ChatGPT’s regular answers, OpenAI says, and advertising has no influence over the core responses you get when you’re not shopping. Advertisers also don’t get access to your other conversations. Given how much scrutiny AI companies are under right now for blending commercial interests into supposedly neutral outputs, that separation is clearly the thing OpenAI most wants people to believe.
On the advertiser side, the tooling is built to remove friction almost entirely. Businesses can build a campaign by describing it in plain language, and OpenAI’s system generates the copy and images. An Ads Manager plugin lets advertisers create, update, and review performance without leaving a chat interface. Shopify merchants in the US get direct integration, and HubSpot is the first CRM partner, meaning marketing teams can plug existing customer data workflows straight into ChatGPT campaigns.
The uncomfortable question nobody’s asking yet
Here’s the thing worth sitting with: an AI that answers your specific, personal questions about a product is a more persuasive salesperson than a banner ad could ever be. A static ad tells you a jacket has a insulated lining. A Sponsored Agent tells you, in a natural, responsive way tailored to the exact question you asked, that yes, this particular jacket will keep you warm on your specific ski trip in your specific climate. That’s a genuinely more useful shopping experience. It’s also a much more effective persuasion machine, wrapped in the trusted, conversational voice people have started treating like a neutral advisor.
OpenAI’s separation promise, that Sponsored Agents don’t touch the “real” ChatGPT, only matters if people actually notice which mode they’re in. Ads that look and feel like the same helpful assistant you use for everything else are, by design, harder to mentally file under “this entity is trying to sell me something” than a Google search ad with a little “Sponsored” tag next to it.
Where this fits in the bigger AI ad land grab
OpenAI isn’t first here. Google has been threading ads into AI Overviews and Gemini for months, and Perplexity has its own sponsored answer format. What makes this one notable is the scale ChatGPT already operates at and how directly it’s now competing with Google for the same advertiser budgets, using a format built specifically to feel less like an ad and more like a helpful expert who happens to work for the brand.
Bottom Line
Sponsored Agents are a smart product and a genuinely useful shopping tool, and also exactly the kind of feature that quietly redraws the line between advice and advertising until most people stop noticing where it moved. Watch how clearly OpenAI actually labels these conversations once the test expands past a handful of advertisers. That’s where the real test of this “separate but trustworthy” promise happens.
