OpenAI plans to start testing visual ads in ChatGPT later this month, placing them in the short wait while an image is being generated. According to OpenAI’s announcement, the test begins in the U.S. with an initial group of advertisers, and the ads will be clearly labeled and kept separate from the image being created.
The company has not given an exact start date. BleepingComputer reports that the format is one of the first new ad types OpenAI is adding as it expands advertising in ChatGPT.
What the visual ads will look like
OpenAI says the ads can use pictures to show product inspiration, how a product is used, or the experiences a service makes possible. In other words, they are meant to work as small visual promotions rather than text links.
The company stresses two promises. The ads will be labeled, and they will stay apart from the generated picture, so an image you ask for is not altered to include a sponsor. OpenAI also reiterated that advertising does not influence the answers ChatGPT gives.
Why image generation is the first stop
Generating an image takes a few seconds, and during that time a person is usually looking at a progress screen rather than reading text. That is our interpretation, not something OpenAI has said, but it helps explain the choice. A visual ad suits a visual task, and the pause gives it a natural place to appear without interrupting an answer.
It is also a placement with a clear boundary. The ad sits beside a process instead of inside a result, which makes the promise of separation easier for users to judge for themselves.
Who will see them
The announcement says the test is for U.S. users. In the coverage we reviewed, OpenAI does not list which ChatGPT plans are included, and reports differ on the details. Android Headlines reports that the test targets free-tier users and that paid subscribers will not see the ads. Until OpenAI confirms the plans, that should be treated as unverified.
Measurement, brand safety and privacy
Alongside the new format, OpenAI is adding tools for advertisers. It is connecting ChatGPT ads to conversion data through Hightouch, Tealium and LiveRamp, and its attribution partners now include AppsFlyer, Adjust, Branch, Triple Whale and Kochava, among others. Attribution means linking a purchase or sign-up back to the ad that may have prompted it.
OpenAI is also working with DoubleVerify and Integral Ad Science on brand suitability, the practice of keeping ads away from content that does not fit. The stated aim is to keep ads out of emotionally vulnerable, sensitive or otherwise unsuitable conversations. According to OpenAI, these independent partners will not get access to private user conversations while they evaluate whether the safeguards work.
That last point matters because ChatGPT ads appear around conversations rather than on a public webpage, where the surrounding content is visible to everyone.
What this means for you
If you generate images in ChatGPT in the U.S., you may start seeing a labeled ad during the wait later this month. The label is how to tell it apart from your result, and OpenAI says the picture you asked for stays untouched.
If you are outside the U.S., nothing changes yet, because the test is limited to that market.
If you pay for ChatGPT, check what you actually see rather than assuming. The plans included have not been confirmed by OpenAI in the material we reviewed.
For small businesses and creators who advertise online, the format is a new option, but no results for it have been published. A test with an initial group of advertisers is not evidence that image ads will perform well, and measurement tools only help once there is data to measure.
What remains unclear
Several practical questions are open. They include the exact start date, which plans see the ads, how often an ad appears per image, and how the brand-suitability checks decide what counts as sensitive in practice. OpenAI has not published performance data for this format.
The useful takeaway for now is simple. Advertising in ChatGPT is reaching a new part of the product, and the promises that matter most, clear labeling and separation from your image, are ones users will be able to judge once the test starts.
