ChatGPT ads can be aimed at you in two ways, and only one of them has much to do with what you type. That’s the short version of a report from 404 Media this week, and of OpenAI’s own help pages. If you use the free or Go version of ChatGPT, here’s how the targeting works, what OpenAI says it won’t do, and what you can switch off.
How ChatGPT ads are targeted
OpenAI’s help page on ads says ChatGPT ads are chosen from the context and intent of your current conversation, plus the ad’s content, its landing page and hints from the advertiser. If personalization is on, some signals from your wider ChatGPT use may count too. When several ads qualify, relevance and the advertiser’s bid decide which one shows.
The second layer is the one that drew attention. Advertisers can upload a list of emails, phone numbers or Google advertising IDs, and OpenAI matches the list to ChatGPT accounts. People who match become eligible to see that advertiser’s ads. OpenAI’s custom audiences page describes the feature, and Gizmodo walked through it on Friday, building on 404 Media’s reporting from Thursday.
This kind of list matching is routine in digital advertising. 404 Media’s framing is that using prompts as a targeting signal is the newer part. That’s the outlet’s characterization, and we haven’t verified it independently. The 404 Media piece is paywalled, so most of what we know about its testing comes through Gizmodo, which says the reporters ran two ad campaigns.
What advertisers can and can’t see
According to OpenAI, advertisers never see your chats, history, memories, name, email, precise location or IP address. They get aggregated numbers such as views and clicks. OpenAI also says conversations aren’t shared with advertisers and that your data isn’t sold.
There’s a subtlety worth stating carefully. Advertisers don’t receive your personal details, but they do supply them. To build a custom audience, they hand OpenAI a file of identifiers. OpenAI says it doesn’t tell advertisers when a match happens, and they can’t pick out individual matched users. Being on a list doesn’t guarantee you’ll see an ad.
Audiences also have a floor. The help page says inclusion targeting and bid adjustments need at least 25,000 matched users, and smaller lists can only be used to exclude people. OpenAI recommends 100,000 or more. Uploaded files are deleted after processing, usually within 24 hours.
How long ChatGPT ads have been around
OpenAI says ad testing began in the U.S. in February 2026. Custom audiences came later: a trade site, PPC Land, reported a limited rollout in May and a wider one in July, but that timeline rests on one outlet. Gizmodo adds that OpenAI has started showing visual ads next to text ones, also from a single report. The audience is large. 404 Media, citing Reuters, puts ChatGPT at about one billion monthly users, so even a small slice of that is a big pool for advertisers.
Who sees ChatGPT ads
ChatGPT ads appear for Free and Go plans. Plus, Pro, Business, Enterprise and Edu plans don’t show them, according to OpenAI. The company says ads are rolling out gradually in select regions, and that personalized ads aren’t initially available in the European Economic Area or Switzerland. Accounts identified as under 18 don’t get ads, and neither do Temporary Chats.
OpenAI also says it keeps ads away from sensitive conversations. Its page says ads won’t run near personal health, mental health or political topics, and that political ads aren’t allowed. Those are the company’s rules. We haven’t tested how well they hold up.
How to limit ChatGPT ads
Start in Settings, then Ad Controls. OpenAI says you can turn off ad personalization, choose whether past chats and memory influence ad selection, hide or report an ad, see why one was shown, and view your ad history.
Turning personalization off doesn’t turn ads off. Contextual ads based on your current chat still appear, as Gizmodo and the OpenAI page both note. If you want no ChatGPT ads, the options are a paid plan, a Temporary Chat, or an “Ads-Free” setting on the free plan. OpenAI says that setting lowers your message limits and restricts some tools, including image generation and deep research.
You can also delete your ads data. OpenAI says that stops it being used to pick ads without affecting your chats, though full removal from its servers can take up to 30 days.
What we don’t know about ChatGPT ads
Several questions are still open. Gizmodo says the personalization opt-out appears to cover ads based on your past chats and activity, and may not stop ads from list-based matching. OpenAI’s documentation we reviewed doesn’t address that either way. Gizmodo also says 404 Media couldn’t confirm whether OpenAI hashes raw identifiers when they’re uploaded, and that OpenAI hadn’t responded to Gizmodo before publication. We couldn’t retrieve an on-record OpenAI comment to 404 Media.
We also don’t know whether users are told when they’re in an advertiser’s audience. The documents say advertisers aren’t told, and nothing we found says users are. A quote from an outside privacy analyst calling the approach “boiling the frog” comes secondhand through another outlet, and we couldn’t check it against the paywalled original.
Why it matters
ChatGPT ads feel different from ads on a search page. People type things into ChatGPT they might not type into a search bar, and the company says it keeps ads away from the most sensitive subjects. We covered the wider regulatory pressure on OpenAI in our piece on the FTC’s look at OpenAI and Anthropic. If you’re choosing between assistants, our comparison of ChatGPT, Claude and Gemini plans covers what each tier costs.
Editor’s take
ChatGPT ads are turning into a normal ad platform, with the usual bid multipliers and audience files. That doesn’t make them sinister, but it does change what “free” means. Your prompt isn’t handed to advertisers, by OpenAI’s account, yet it can still shape what you’re shown. The practical move is small: open Ad Controls, decide what you’re comfortable with, and use Temporary Chat for anything you’d rather keep out of the mix. And if the opt-out matters to you, ask OpenAI directly whether it covers email and phone matching. Right now, its documents don’t say.
