OpenAI to Test Image Ads Inside ChatGPT's Picture Generator

OpenAI said Monday it will begin testing a new ad format that inserts images of sponsored products and services into ChatGPT, with the trial starting in the US later this month. The ads will initially appear when a user generates an image, meaning the sponsored placements sit alongside or within visual output rather than in the text conversation. The company framed the rollout as a test, and the wording of its announcement suggests the format could spread to other parts of the product if it performs.

The move extends an advertising business OpenAI opened in February, when it first put ads inside ChatGPT. That earlier phase kept commercial placements away from the image generator, according to the company's own description of the launch. The new format is the first time OpenAI has attached advertising to visual output, a surface where sponsored imagery is harder to separate from the answer a user actually asked for. It also puts OpenAI's ad inventory closer to the kind of product-placement real estate that search engines and social platforms have sold for years.

For OpenAI, the calculus is straightforward: ChatGPT has a very large user base and inference costs are heavy, so advertising offers a revenue line that does not depend on subscriptions or API usage. The company has been under pressure to show that consumer scale converts into durable income as it funds model training and compute. Image generation is expensive per request, which makes it an obvious candidate for offsetting some of that cost with sponsored content. Whether advertisers will pay enough to matter is a separate question the test is designed to answer.

The risk sits with users and with the output itself. When a sponsored image appears in a generated picture, the line between what the model produced and what a brand paid to show can blur, particularly if the placement is not clearly labeled. Regulators in the US have been scrutinizing how AI systems disclose commercial relationships, and the Federal Trade Commission has taken an interest in native advertising that mimics organic content. OpenAI will need to make the labeling obvious enough to survive that scrutiny while keeping the format attractive to buyers.

OpenAI has not said how many ads users will see, how they will be targeted, or whether the format will reach free and paid tiers alike. Those details will determine whether this reads as a modest test or the start of a broad ad load across ChatGPT's visual features. The company also has not disclosed pricing for the placements or named launch advertisers, which makes it hard to judge how serious the bet is. Watch the labeling first, then the volume.