ChatGPT’s new visual ad format will first appear while a user waits for an image to generate, a moment with little or no shopping intent behind it. The placement shows ChatGPT Ads moving past the idea that an ad has to match what a user asked.
OpenAI announced the format on Monday alongside its new measurement partners. Eric Seufert, the analyst who writes Mobile Dev Memo, focused his analysis on the format, and his reading explains what the placement changes.
What OpenAI Is Testing
The OpenAI announcement describes images showing product inspiration, product usage or the experiences a product makes possible. The first test runs during image generation in ChatGPT.
Ads will be labeled and kept separate from the image being created. OpenAI says advertising does not influence the answers ChatGPT provides.
Testing begins later this month in the US with an initial group of advertisers. OpenAI has not named those advertisers or given a date for a wider rollout. It puts ChatGPT’s reach at 1.2 billion people each week, though the ads reach only part of that audience: OpenAI’s help documentation says ads may appear for users on the Free and Go plans, while Plus, Pro, Business, Enterprise and Edu accounts will not have ads, and accounts identified as belonging to under-18s are not shown ads.
A Display Ad Without a Commercial Question
Eric Seufert calls the new unit a display ad. His point is about the setting. A user waiting on a generated picture has usually not asked anything commercial, so there is no subject in the conversation for an ad to match.
The usual approach works the other way. Asad Awan, who leads monetization product and engineering at OpenAI, has described a keyword system as one that maps a trip to Yosemite to a hotel or a flight.
Seufert argues that chatbot advertising has an either/or advantage. When a conversation has a commercial subject, the ad can be anchored to it. When it does not, the ad can be anchored to the conversion data advertisers already send to platforms through their pixels and Conversions API (CAPI) connections.
The same announcement widened that second route. Hightouch, Tealium and LiveRamp now help advertisers send conversion data from their existing systems into ChatGPT Ads. Awan has also said ChatGPT could choose ads by what it remembers about a user, not only by the conversation on screen.
OpenAI Described the Placement Last Week
The test matches what Awan told Seufert on the Mobile Dev Memo podcast, which The Keyword covered on 1 October. Of image generation, Awan said: “I am waiting for the image to get generated, that is a really good place for us to experiment.” In that moment, he said, “I am using the product like a tool not to answer a specific thing.”
He also said the idea applies to “other places which are tool style usage of the product like write an email for me.” An email draft, like an image, is a task rather than a question. A contextual ad matched to it might be an email writing tool, which Awan said “is not very interesting for the user.”
That line is the clearest sign of where the format could go next. Image generation is the first tool-style surface to carry an ad test, and Awan named writing tasks as another.
Ads That Need Not Be Text
Seufert made the format case in November 2025, before ChatGPT carried any ads. He argued then that chatbot output would not stay text, since most chatbots already take in and produce more than words.
… a consumer’s chatbot engagement may more resemble YouTube than Google Search, and so should the underlying business model.
In his view, a chatbot ad can look exactly like an ad in a feed or in a streaming video. ChatGPT has tried larger images before, with a bigger image and CTA button reported in May, and job listings in July pointed to image, video and interactive formats. Both concerned what the ad looks like. The image generation test changes where it appears, in a moment that does not depend on an answer at all.
Google’s newest AI Mode formats still follow the query. Its Conversational Discovery ads, announced in May, use Gemini to write ad creative for each Search query.
What Changes for Advertisers
For advertisers, the test changes what a ChatGPT campaign is built from. A text ad matched to a prompt depends on copy that fits the question. A visual ad shown during image generation depends on imagery of the product in use, and on conversion data good enough for ChatGPT to choose a relevant user without a commercial prompt.
Recap
What is ChatGPT’s new visual ad format?
It is an ad format built on images showing product inspiration, product usage or the experiences a product makes possible. OpenAI will first test it during image generation in ChatGPT, labeled and kept separate from the image being created, starting later this month in the US with an initial group of advertisers.
How can an ad be targeted when a user has not asked a commercial question?
Eric Seufert argues that when a conversation has no commercial subject, the ad can be anchored to conversion data advertisers send through pixels and the Conversions API. OpenAI added Hightouch, Tealium and LiveRamp as routes for that data, and Asad Awan has described ads chosen by what ChatGPT remembers about a user.
What does the test change for advertisers?
ChatGPT ads no longer have to match the question a user asked. Campaigns for the visual format depend on imagery of the product in use and on conversion data strong enough to pick a relevant user, and Awan has named writing tasks as another surface where the approach could apply.