The ad test is confirmed; the visual format is not
OpenAI is testing ads in ChatGPT. The company has presented the initiative as a way to support access to the service and has outlined general principles for its advertising, including clear identification and a separation between ads and the assistant’s responses. That announcement confirms that advertising tests are underway, but it does not, by itself, verify every format or placement that might be included.
The development reported in connection with image creation is more specific: visual ads that would appear during the process of generating an image. Some descriptions present them as items separate from the generated result and labeled as advertising. However, the official information currently available about ads in ChatGPT does not confirm this particular integration. For now, the format should be treated as provisional information, not as a fully detailed feature officially announced by OpenAI.
The distinction matters. OpenAI testing ads in ChatGPT is supported by the company’s own statements. A separate claim that an ad will appear alongside a generated image, on an intermediate screen or after image creation is complete is not the same thing. The available information does not establish with confidence which of those experiences is being tested.
Where might an ad appear—and why the timing is still unclear
The description that has circulated places the ad in the image-creation flow, but that phrase is not precise enough to define the experience. It could mean an item shown while the system is working, a module on a waiting screen, or a space that appears once the image is ready. No primary information available here establishes which of these would be the case.
The frequency, size and duration of such ads are also unknown, as is whether they would appear with every image request. It is not possible to conclude that they would be integrated into the image, change the result or always appear beside it. The central question is whether an ad would be shown as a separate element and, if so, at what point in the interface.
For someone using the image generator, those scenarios are not interchangeable. An ad on a waiting screen competes for attention while a request is being processed. One placed beside the result could be mistaken for part of the creation experience if it is not clearly distinguished. An item incorporated into the image would raise different concerns. None of these scenarios should be treated as confirmed without a more precise product description.
How the experience would differ depending on when an ad appears
These scenarios illustrate which details are missing; they do not describe a confirmed interface.
| Possible placement | What the user might see | What remains unknown |
|---|---|---|
| While waiting | A separate item while the request is being processed | How long it remains visible, how often it appears and whether it can be dismissed |
| Beside the result | Advertising next to an image that has already been generated | How it is distinguished visually and whether it shares a screen or module with the image |
| Inside the image | An item that could be mistaken for generated content | There is no confirmation that OpenAI will use this format |
Initial reach: the United States is an indication, not a complete rollout plan
Information about the reported test places its start in the United States. Even so, the official material currently available about advertising in ChatGPT does not confirm that a visual format linked to image generation would be limited to that country, nor does it provide a launch schedule. The geographic scope should therefore be understood as provisional, not as a defined availability plan for all users in that market.
The Free and Go plans have also been mentioned in connection with ad testing in ChatGPT. That provides context about who might encounter advertising in the service, but it does not establish that every plan would receive the same visual format or that an image-related test would apply to every account included in those tiers.
There is not enough information to say when a visual-ad test would begin, how long it would last, what share of people would take part or whether it would expand to other countries. The existence of a general advertising test also does not mean that the specific image-generation format is active for the public.
How to tell advertising apart from generated content
OpenAI has described clear identification as a principle for ads in ChatGPT. That is a relevant reference point for evaluating any future format, but it is not a design specification for visual ads during image generation. The wording of a label, its placement and size, and the visual elements that would separate advertising from the image produced by the system are not known.
In practice, a readable label placed close to the ad would help prevent confusion. It would also matter whether the ad stayed separate from the image file and from the controls used to edit or download it. These are useful criteria for assessing a design if one is introduced; they are not features that can already be attributed to the test. Without a screenshot or an official product explanation, it is not possible to verify how that separation would work.
The distinction also depends on what counts as an ad. A commercial recommendation in a response, a sponsored module and an image generated at a user’s request are different things. A future interface should make it possible to recognize which element is advertising and which is part of the requested result. The available public information does not explain how that interaction would work in this case.
A quick check for users
These are practical criteria for interpreting an interface, not a description of controls already available in the visual-ad test.
- 01Look for an explicit label identifying the item as advertising, rather than only a brand name or logo.
- 02Check whether the ad is separate from the generated image and from the options for modifying it.
- 03Do not interpret a product or brand inside an image as advertising without clear evidence that it is an ad.
- 04If controls for hiding or managing ads are provided, check what they do and which preferences they let you change.
Measurement, personalization and controls: questions that remain open
OpenAI’s help and policy pages explain general aspects of its advertising, including eligibility, personalization, privacy and controls. That material helps explain the broader framework for advertising in ChatGPT, but it does not confirm which data would be used to select ads alongside images, which signals would be excluded or what specific options a user would have in that experience.
Measurement is another open question. There is no detail on how impressions or interactions would be counted in a format tied to image generation, or whether advertisers would have attribution tools. Nor is there enough information to say that such tools are already available for the format described. A general advertising policy is not enough to infer how a specific test would be measured.
Before the real-world impact can be assessed, answers are still needed about the exact placement, labeling, separation from the result, selection criteria, data use, user controls and measurement. Until OpenAI publishes those details or the feature can be checked directly, the most careful conclusion is limited: the company has confirmed advertising tests in ChatGPT, while visual ads during image creation still lack sufficient official confirmation.
Open questions
- There is no official confirmation available here about the exact placement of an ad in the image-creation flow.
- It has not been verified how advertising would be identified visually or separated from the generated image.
- References to the United States and certain plans do not establish the full reach or schedule of a visual-ad test.
- Data used for personalization, format-specific user controls, and measurement or attribution methods have not been detailed.
Keep exploring
Sources consulted
Corrections and transparency
If you spot incorrect or outdated information, send us a correction with the page and source we should review.
Submit a correction