The new format
OpenAI said on Monday that it is adding visual display ads to ChatGPT, shown alongside the images a user asks the model to generate. The format rolls out later this month in the United States only, with an initial group of advertisers, and appears for free and lower-cost subscription users.
The ads can use imagery to show a product, how it is used, or an experience associated with a service — a more visual placement than anything OpenAI has offered before. The company says they will be clearly labelled and that they “won’t influence the answers ChatGPT provides”.
What this slot actually is
Image generation takes several seconds. Until now that was dead time. Putting a sponsored unit under the progress bar turns the wait into an impression, and it does so at the moment a user has just described, in their own words, something they want to see.

That is a sharper targeting signal than a search query, and it arrives without the user having typed anything into an ad product. A request to generate a picture of a kitchen renovation is, commercially, a kitchen renovation intent. Whether OpenAI uses the prompt to select the ad is the question the announcement does not settle.
The rest of the ad stack
This is not OpenAI’s first ad product. It introduced a text-based format earlier in 2026, a single sponsored unit at the bottom of ChatGPT answers, and expanded that format to India in August. The visual placement sits on top of that, extending advertising from the text surface to the image surface.

The commercial logic is straightforward. ChatGPT’s free tier is enormous and does not pay for itself, and advertising is the only mechanism at that scale that does not depend on converting those users to a subscription. The strategic question is how much advertising the product can carry before the thing people value about it — an assistant that answers rather than sells — starts to erode.
ChatGPT’s reach makes the placement unusually valuable on its own terms: OpenAI has said the assistant has more than a billion weekly users, and image generation is among the features free accounts use most.
What to watch is the labelling in practice and whether the rollout stays inside the United States. The EU’s rules on transparency and on profiling-based advertising apply differently, and OpenAI has already shown, with its EU-only text watermarking announced the same day, that it is willing to run different products in different jurisdictions.