People creating an image with ChatGPT may soon see visual ads while it is being generated. On October 5, OpenAI announced a new ad format that it plans to test in the United States later in October. For users, the central question is whether they can clearly distinguish a paid product suggestion from their own image and the assistant’s response.
Key takeaways
- The new visual ads will begin as a US test with an initial group of advertisers, rather than a worldwide launch.
- OpenAI plans to label the ads and keep them separate from the generated image. The company says advertising does not influence responses.
- It is also expanding measurement of purchases and other business outcomes. However, an attributed purchase does not establish an additional effect caused by advertising.
- Existing ad rules distinguish Free and Go from ad-free subscriptions. No German launch date has been announced for the new format.
Ads beside the creation, rather than inside it
The announcement initially concerns where ads appear. The new format will present products and services visually, including scenes showing how they might be used. OpenAI identifies image generation as the first testing area. This puts an advertising offer into a moment when users might already be thinking about colors, furnishings, clothing, or other visible characteristics. The announcement does not fully specify which topics will trigger ads.
A hypothetical example makes the distinction clearer: Someone requesting an illustration of a living room is still asking for their own image. A furniture ad displayed beside it would be an additional paid placement. This example describes a possible use case, rather than a confirmed test campaign. In particular, the announcement does not imply that brands will automatically be inserted into the generated image.
The Decoder reports that the ads will appear as a carousel or a row of images below the image being created. The manufacturer’s stated commitment is to label ads and keep them separate from the image. The test will show how recognizable that boundary is on small smartphone screens. An ad label helps only when the visual relationship is clear as well.
This is a different development from the previously covered sponsored agents in ChatGPT (German article). This announcement concerns an ad format alongside a creative task. The planned test is neither a new model version nor evidence that image quality will improve.
Why a purchase alone does not prove an ad worked
The second half of the announcement addresses businesses. Interfaces and partners are intended to help them better understand what happens after an ad. A conversion is a desired action, such as a purchase, registration, or inquiry. This information helps retailers evaluate advertising budgets instead of looking only at clicks.
OpenAI’s accompanying measurement report describes a tracking pixel and a Conversions API, among other tools. The pixel records events on a website; the API allows events to be sent from existing systems. The company names integrations with Hightouch, Tealium, and LiveRamp. These are routes for business data to reach the advertising platform, rather than an announced release of private chat histories to advertisers.
The difficult part is attribution. If someone sees an ad and later makes a purchase, a measurement system may attribute that purchase to a campaign. This does not yet answer whether the person would have bought the product without the ad. OpenAI therefore distinguishes attribution, which assigns credit, from measuring additional impact. For the latter, it is exploring geographic comparison experiments with partners.
For a small retailer, this suggests a practical rule: Start a new campaign with a clear objective and a manageable budget. Sales, costs, and an appropriate comparison need to be evaluated together. An impressive figure from someone else’s campaign cannot replace examining your own customers. The early partner results published by OpenAI are case studies, rather than a guarantee of returns from the new visual format.
What users can check now
For users in Germany, availability is the first question. The new announcement refers only to a US test later in October. Seeing existing ads in a German account therefore does not mean these visual ads have also been enabled. Likewise, the absence of an ad area does not indicate an image generation error. OpenAI has not provided a specific German launch date for this format.
The general advertising help page lists Free and Go as plans that may carry ads. Plus, Pro, Business, Enterprise, and Edu are supposed to remain ad-free. This describes existing plan rules; the new format announcement does not provide a detailed rollout schedule for every account. Anyone considering changing their subscription solely because of this news should first check whether the test affects them at all.
Where ad controls are available, the help page says users can manage personalization and advertising data. Ads can be hidden or reported. Turning off personalization does not remove advertising, however: The current conversation’s context can still inform ad selection. According to the help page, personalized ads are not initially available in the European Economic Area or Switzerland. Regional differences should therefore not be inferred from a US screenshot.
OpenAI says advertisers receive aggregated performance information rather than private conversations. The new projects evaluating the advertising environment are also intended to operate without access to real private user conversations. These statements describe the announced safeguards. They are distinct from the broader question of what data a retailer collects on its own website after someone clicks an ad.
The critical boundary is on the screen
Visual ads can make products easier to understand than a short line of text. At the same time, they compete with the creative result for attention. The planned test should therefore be judged by more than the purchases it generates. Users need to be able to distinguish their request, the generated image, and the paid placement at all times. Retailers need measurement that can identify additional demand. Only when both questions are answered convincingly will the format amount to more than another advertising slot in a chat.