OpenAI tests ChatGPT ads that open branded AI conversations
OpenAI is testing Sponsored Agent ads that turn a click into a branded conversation inside ChatGPT. The format creates a new commerce surface, but gives retailers a harder choice between conversational convenience and control of the customer journey.
OpenAI is testing a ChatGPT ad format that opens a branded AI conversation instead of sending a user to an advertiser's website. The Sponsored Agent pilot lets selected advertisers attach a business agent to an ad, turning a call to action such as "Chat with us" into an in-product exchange.
Wayfair has confirmed that it is participating at limited scale. The retailer is using the pilot to explore whether conversational commerce can help shoppers make more informed decisions, while applying guardrails around product accuracy, transparency, and handoffs to customer service.
The test is still early, and its commercial value remains unproven. Even so, it sharpens an important question for marketers: if the ad becomes the start of a service interaction, who owns the experience after the click?
Table of contents
Jump to each section:
- How the Sponsored Agent format works
- Why click-to-chat changes the ad unit
- The ownership tradeoff for retailers
- What marketers should know about conversational ads
How the Sponsored Agent format works
ChatGPT's existing ad experience places sponsored units beneath organic responses. The new format adds another layer: an advertiser can connect a branded business agent to the placement, so a click opens a dedicated conversation within ChatGPT rather than a browser tab.
That makes the agent part of the media experience. It can help a user continue researching, narrow options, or move toward a service or sales handoff without requiring an immediate channel change. For Wayfair, the pilot is framed around helping shoppers make more confident decisions, not simply generating a visit to a product page.
The distinction is small in interface terms but meaningful in marketing terms. A conventional ad earns attention and then transfers the user elsewhere. A sponsored agent tries to hold the interaction, answer the next question, and qualify intent within the same environment.
The click is no longer merely a handoff. It becomes permission to continue the conversation.

Why click-to-chat changes the ad unit
Most digital ad formats are built around interruption or redirection. Search ads lead to landing pages, social ads lead to shops or forms, and display placements lead to branded destinations. Sponsored Agent ads point toward a different model, where the advertiser's immediate task is to sustain a useful exchange.
That changes what creative has to do. The ad still needs to earn a click, but the experience that follows must also understand questions, present accurate product information, and know when to escalate. Copy, product data, service design, and agent behavior therefore become parts of the same advertising system.
The common assumption is that conversational ads are simply more interactive placements. In practice, the interaction may make them closer to a lightweight sales and service channel. That means a media team cannot evaluate the format only through impressions and click-through behavior. The quality of the conversation and the outcome of the handoff matter too.
Every major ad platform eventually develops a format that reflects its native behavior. ChatGPT's most natural behavior is not scrolling or browsing. It is asking, refining, and responding. Sponsored Agent ads are an attempt to make advertising follow that rhythm.
The ownership tradeoff for retailers
Keeping a shopper inside ChatGPT can reduce friction, especially when the person is still comparing products or clarifying a need. It can also delay the moment when the advertiser gains a direct site visit, first-party behavioral data, or control over the surrounding experience.
For retailers, that tradeoff is unusually sharp. A useful agent could help a shopper move closer to a decision before reaching the retailer's site. But the more value the conversation delivers inside ChatGPT, the more important it becomes to define where the agent stops, how product claims are governed, and when the user should move into an owned channel.
Wayfair's guardrails reveal the real work behind the format. Product accuracy is not just a customer-experience concern when an AI agent represents the brand. It is also a media-quality issue. Transparency affects trust, and service handoffs determine whether conversational interest becomes a resolvable customer need.
Convenience can shorten the journey while making ownership of that journey less clear.
The format therefore creates a strategic tension between reach and control. Brands may gain access to high-intent conversations, but they must decide which parts of discovery, qualification, and service they are comfortable conducting on another company's platform.
What marketers should know about conversational ads
The Sponsored Agent pilot should be treated as a test of interaction design as much as a test of media performance. Four questions deserve attention before the format earns a permanent place in a plan.
Define the job of the conversation. A branded agent needs a clear role, such as helping users compare options or identify the right product. A vague mandate risks producing a longer interaction without improving the decision.
Measure beyond the initial click. The meaningful outcome may be a qualified recommendation, a transfer to customer service, or a visit to a relevant product page. Click volume alone cannot show whether the agent made the journey better.
Set boundaries for claims and handoffs. Wayfair's focus on accuracy, transparency, and service escalation is a useful signal. The agent needs to know both what it can answer and when a person or owned channel should take over.
Protect the value of owned channels. A smooth in-platform experience can be useful without becoming the entire journey. Marketers should understand what data, context, and customer relationship they gain or give up at each transition.
OpenAI is still looking for an ad format that feels native to an AI assistant, and this pilot may change substantially as advertisers and users respond. Its significance is less about replacing landing pages than about expanding what an ad is expected to do.
As AI platforms move closer to product discovery and commerce, the line between media, software, and customer service will keep thinning. The brands best prepared for that shift will not be those that generate the most conversational copy. They will be the ones that can design a useful exchange, govern it carefully, and connect it to a customer journey they can still understand.

