Where ChatGPT ads sit and what data comes back

The mechanic that matters most on ChatGPT Ads is where the ad sits. An advertisement appears below a ChatGPT response when OpenAI determines that there is a relevant sponsored product or service. It will be clearly labelled as sponsored and will be visually separated from ChatGPT’s answer. OpenAI’s description is explicit that the advertisement sits alongside the response, rather than being part of the AI-generated answer itself.

That one design decision settles a question brands have been asking since the format was rumoured. Buying here does not buy a mention inside the model’s reply. The answer stays the answer and the ad is a separate labelled unit next to it, which puts this closer to a marked placement than to anything that shapes what the assistant says.

Targeting is described by OpenAI in terms of what it calls high-context moments, points where a consumer’s decision is forming, based on what users are actively trying to figure out. That is OpenAI’s own framing rather than a documented targeting spec, and the article does not set out what controls an advertiser gets over it. Two exclusions are stated: ads will not be eligible to appear alongside sensitive or regulated topics, including mental health and politics.

Eligibility on the user side is narrow by design. Free and Go tier users see ads. Plus, Pro, Business, Enterprise and Education users do not, and users under 18 will not see ads at all. Users can dismiss an ad, give feedback on it, understand why it was shown, manage advertising personalisation and clear the data used for advertising. No approval or review process for advertisers is mentioned in the article, which is not the same as saying there is none.

Measurement is where an Indian operator should set expectations early. Advertisers receive aggregate campaign-performance information, such as impressions and clicks. Advertisers will not receive access to individual conversations, and OpenAI does not sell user data to advertisers. That makes the reporting thinner than the audience and placement detail a brand is used to pulling off Meta or Google. Plan to measure this the way you would measure an offline placement: one clean dedicated landing path, a deliberate holdout window where you run nothing, and a willingness to judge the surface on what moves in your own site and marketplace numbers rather than on what the ad platform hands back.