Advertisers can now hand OpenAI a list of email addresses and phone numbers, and ChatGPT will show ads to the people on it. Your prompts feed the targeting too. That’s how ChatGPT ad targeting works right now, and most of the app’s 1.2 billion weekly users have no idea.
How advertisers find you inside ChatGPT
404 Media ran a real ad campaign inside ChatGPT to test this, and what they found is wild. OpenAI lets advertisers build something called a custom audience. You upload a list of email addresses, phone numbers, or Android advertising IDs, and OpenAI matches them against real ChatGPT accounts. If your email is on that list, you can get targeted. The advertiser can upload raw data or a hashed version, which is a scrambled copy meant to hide the original.
Here’s the detail that raised eyebrows. Meta hashes this kind of data by default when advertisers upload contact lists. 404 Media asked OpenAI whether it does the same and got no answer, just a reminder that advertisers can upload both raw and hashed versions. OpenAI did say uploaded files get deleted after processing, usually within 24 hours.
There’s a catch that works in your favor, sort of. A list only counts for targeting once it matches at least 25,000 users. Below that, your data can only be used to exclude you from campaigns. Advertisers also never see which individuals matched.
The other targeting tool is sneakier. Advertisers add “context hints” describing the topics where their ads should appear. So if a company uploads “people asking about privacy” as a hint, ChatGPT decides whose prompts look like privacy questions. Zach Edwards, co-founder of DecryptAds, told 404 Media that OpenAI is “boiling the frog” and hoping users won’t notice the privacy implications.
How ChatGPT ad targeting picks your ads
Here’s the official breakdown, straight from OpenAI’s help center. Some of it you’d expect. Some of it surprised me.
- Your current chat thread. Always. This one has no off switch.
- General context, like your rough location and language.
- Past chats and memory, if personalization is on.
- How you’ve interacted with previous ads, like hiding or clicking them.
Now the part OpenAI wants you to know: advertisers never receive your chats, your email, your IP address, or sensitive categories like health and political topics. What they actually get is aggregate stuff, meaning view and click counts. One exception worth remembering: if you message a company through their ad, they see whatever you send them.
Ads also don’t run next to conversations about personal health, mental health, or politics. And in the European Economic Area and Switzerland, personalized ads aren’t enabled at all for now.
The rollout has been fast, though. Ads hit the US in February, then the UK, Mexico, Brazil, Japan, and South Korea by August. This month OpenAI started placing ads right next to image generation results, which means you can now get a sponsored card while you’re waiting for a picture of your cat.
Want proof this happens at scale? A third-party site called ChatGPT Ad Library probes the chatbot with everyday prompts and records every sponsored card that comes back. It has captured over 43,000 ads across more than 415,000 placements.
Turn off ChatGPT ad personalization in five minutes
The controls for ChatGPT ad targeting actually exist, which is more than most companies manage. The problem? Almost nobody knows where they are.
- Open Settings in ChatGPT, then pick Ad Controls. On the web you can jump straight to chatgpt.com/#settings/Advertising.
- Turn off Personalize ads. Ads won’t disappear, but they’ll be chosen from your current chat only, not your history.
- Turn off Past chats and memory. This stops saved memories and older conversations from feeding ad selection. If you only do one thing, do this one.
- Tap View History and Topics, then Delete ads data. This wipes the interests ChatGPT has built up about you. It stops being used for ads right away, though full removal from OpenAI’s systems can take up to 30 days.
- When an ad appears, tap the three-dot menu on it. You can hide that specific ad, report it, or pick About this ad to see why you got it.
One honest note: turning personalization off changes which ads you see, not whether you see them. The ads keep coming on Free and Go plans either way. And after a big ChatGPT app update, double-check these toggles survived it, because this settings panel is still moving around.
The temporary chat trick
This is my favorite loophole, and it’s sitting in OpenAI’s own FAQ: Temporary Chats don’t show ads at all. Nothing gets saved to your history, nothing feeds memory, and ChatGPT ad targeting has nothing to work with. You keep unlimited text chats too. For anything sensitive, health questions, job hunting, money stuff, a temporary chat is the private route. Regular chats still have their place, but the ads come along for the ride on Free and Go.
If you want the ads gone completely
Two real options, and one fake one. The fake one is turning off personalization, which as we covered doesn’t remove a single ad. Here’s what actually works:
| Plan | Price | Ads? | The deal |
|---|---|---|---|
| Free | $0 | Yes | Can switch to Ads-Free mode: fewer messages, no image generation or deep research |
| Go | $8/mo | Yes | No Ads-Free option; drop to Free or upgrade instead |
| Plus | $20/mo | No | Full features, zero ads |
| Pro | From $100/mo | No | Max everything, zero ads |
| Business / Enterprise / Edu | Varies | No | Work and school accounts never see ads |
Ads-Free mode deserves a closer look because it’s free. The fine print: fewer messages per day, no image generation, no deep research. It’s also a Free-plan-only feature, so Go subscribers have to drop down to Free or move up to Plus. Switching back takes one tap on a “Show ads” button whenever you hit the lower limits.
My take: if ads genuinely bother you and you use ChatGPT daily, Plus at $20 is the clean fix. If you’re a light user, Ads-Free costs nothing in money, just limits. And if you mostly want privacy rather than an empty feed, the personalization toggles plus temporary chats get you 90% of the way for free.
Why this suddenly matters
Your prompts feel private because you type them like you’d whisper something to a friend. ChatGPT ad targeting breaks that illusion in two ways. First, humans can review chats, which 404 Media documented in its Project Lily investigation. Second, those same prompts now decide which ads find you. Between this, invisible watermarks rolling out in Europe, and ads creeping into image results, the quiet phase is clearly over.
It also fits a bigger pattern. We covered how Meta Muse lets you see and delete what it knows about you, and how AI shopping recommendations quietly cost people more money. Every AI company is figuring out monetization right now, and the default is always “use everything we have until someone complains.”
The bottom line
Open the app, go to Settings, then Ad Controls. Flip off Personalize ads and Past chats and memory, then hit Delete ads data. Thirty seconds, no downsides, and your prompts stop feeding the ad machine. That’s the whole move.
