OpenAI to test image-based ads in ChatGPT during image generation in the US
In short
OpenAI said it will begin testing a new visual advertising format inside ChatGPT in the second half of October 2026, initially with some advertisers in the United States. The ads appear while a user's image is being generated, below the image in progress, and are labelled as advertising; OpenAI said they do not affect ChatGPT's answers. The company also announced expanded ad-measurement integrations and brand-suitability checks with outside partners.
Read the full story 2 min read
OpenAI will begin testing a new visual advertising format in ChatGPT during image generation in the second half of October 2026, initially for users in the United States and with some advertisers taking part, according to the company's announcement as reported by Gigazine, Impress Watch and Mynavi News. Mynavi News dated the announcement to October 5 US time. [ 1 , 2 , 3 ]
The format shows image-based advertising while a user's image is being generated, displayed below the image in progress. OpenAI said the ads are marked as advertising — Mynavi News said an “Ad” label is shown — and are kept separate from both the generated image and the chatbot's answers. The company said the ads do not influence ChatGPT's replies. [ 1 , 2 , 3 ]
OpenAI also described how it decides where ads may appear. According to Gigazine and Mynavi News, the system assesses whether a conversation is suitable for advertising, aiming to exclude users in a psychologically vulnerable state, sensitive contexts and other unsuitable situations. Advertisers that meet the criteria can use an “exclusion phrases” function to apply their own placement requirements, the three outlets reported. [ 1 , 2 , 3 ]
For advertisers, OpenAI is expanding measurement. Impress Watch and Mynavi News said integrations with Hightouch, Tealium and LiveRamp are intended to make it easier to send conversion data from existing systems into ChatGPT Ads. Both outlets also listed attribution partners including AppsFlyer and Adjust, providing conversion APIs, reporting and click attribution. [ 2 , 3 ]
On brand suitability, Impress Watch and Mynavi News reported that OpenAI is developing an assessment programme with DoubleVerify and Integral Ad Science. Under the plan, the independent partners evaluate how brand-safety standards are applied in a controlled test environment, and OpenAI said they do not access private user conversations. [ 2 , 3 ]
Mynavi News placed the announcement in the context of ChatGPT's existing advertising: it said ads began test display in the US on February 9 for logged-in adult users on the Free and Go plans, and that ChatGPT ads reached an annualised revenue run rate of $1 billion in under 200 days and are available in more than 40 countries. Gigazine noted that ChatGPT already carries advertising and that in September 2026 a test began offering a dedicated AI chat to users who click an ad. Neither detail appeared in the other two documents. [ 1 , 3 ]
Why it matters
The test puts advertising directly inside a generative AI interface used by a mass consumer audience, in a format tied to the moment a user is creating an image. OpenAI's stated guardrails — excluding vulnerable or sensitive contexts and letting advertisers set exclusion phrases — will be watched as a template for how ads are kept apart from AI responses. The company says ChatGPT ads reached a $1 billion annualised revenue run rate in under 200 days and are available in more than 40 countries, making the format's rollout relevant beyond the US test.
Key facts
- OpenAI announced it will test visual ads inside ChatGPT during image generation, starting in the second half of October 2026 with a limited set of advertisers in the United States. [ 1 , 2 , 3 ]
- The ads are shown below the image being generated and are labelled as advertising, separately from the generated image. [ 1 , 3 ]
- OpenAI said ads do not influence ChatGPT's answers and are displayed separately from them. [ 1 , 2 , 3 ]
- OpenAI said its systems assess whether a conversation is suitable for ads, excluding users in a psychologically vulnerable state, sensitive contexts and other unsuitable contexts. [ 1 , 3 ]
- Eligible advertisers get an “exclusion phrases” feature to apply their own placement requirements. [ 1 , 2 , 3 ]
- OpenAI said it is expanding measurement integrations with partners including Hightouch, Tealium, LiveRamp, AppsFlyer and Adjust. [ 2 , 3 ]
- DoubleVerify and Integral Ad Science are developing a programme to assess compliance with brand suitability standards, which OpenAI said runs in a controlled test environment without access to private user conversations. [ 2 , 3 ]
- According to Mynavi News, ChatGPT ads reached an annualised revenue run rate of $1 billion in under 200 days and are available in more than 40 countries. [ 3 ]
Confirmed by several sources
- OpenAI will test a visual ad format shown during image generation, starting in the second half of October 2026 in the United States with a limited group of advertisers. [ 1 , 2 , 3 ]
- The ads are identified as advertising and displayed separately from the image the user is generating. [ 1 , 3 ]
- OpenAI said the ads do not affect ChatGPT's answers. [ 1 , 2 , 3 ]
- Advertisers that qualify can use an “exclusion phrases” function to apply their own placement requirements. [ 1 , 2 , 3 ]
- OpenAI is expanding ad measurement through integrations with outside partners. [ 2 , 3 ]
- A brand suitability assessment programme is being developed with DoubleVerify and Integral Ad Science, without access to private user conversations. [ 2 , 3 ]
Still unclear
- The exact start date of the US test within the second half of October 2026. The documents say only “the second half of October”; no specific date is given.
- How many advertisers will take part initially and which brands they are. The documents say only that some or a limited number of advertisers will participate.
- Whether the format will be expanded outside the United States. All three documents describe the test as US-only, with no statement about other markets.
- How ads are targeted or selected for a given user. The documents describe placement context and safety checks but not the targeting mechanism.
- The $1 billion annualised revenue run rate and 40-plus-country availability of ChatGPT ads. This figure appears only in Mynavi News, a single source.
- The September 2026 test of a dedicated AI chat for users who click an ad. This is reported only by Gigazine and is not mentioned in the other documents.
What local media are saying
Timeline, local time
- OpenAI announces the visual ad test for ChatGPT's image generation, according to Mynavi News, which dates the announcement to October 5 US time. [ 3 ]
- Gigazine publishes its report on the planned image-based advertising in ChatGPT. [ 1 ]
- Impress Watch reports the test, listing measurement and brand-safety partners. [ 2 ]
- Mynavi News publishes its report, adding that the test starts in the second half of October for US users and citing ad revenue figures. [ 3 ]