I asked ChatGPT a technical question, and minutes later an ad for the exact tool I mentioned slid into the chat. You felt the nudge too — a brand showing up like it had followed your trail. I wanted to know how personal that nudge really was.
My chat showed an ad for a company I’d never searched for in my browser.
Here’s the short version: OpenAI is selling advertising inventory inside ChatGPT, and advertisers can aim those ads at specific people. I’m not talking about broad categories like “small business owners” — the targeting can be tied to individual identifiers tied to your account.
404 Media ran two campaigns and walked the process from the advertiser side. Their report shows advertisers can upload lists of identifiers — raw emails, phone numbers, hashed versions, even the Android advertising ID assigned to a device — and OpenAI checks for matches inside ChatGPT. If a match exists, that account becomes eligible to see the ad.
How does ChatGPT target ads to users?
OpenAI’s ad interface accepts multiple match keys. Advertisers can submit a CSV containing emails or phone numbers, hashed values, or Android advertising IDs. Per OpenAI documentation, the system will run a match and hide the advertiser’s identity when a hit occurs, so it won’t hand you someone’s ChatGPT profile like a phone book.
An advertiser uploading a list expects the platform to protect that file.
That’s where the questions begin. 404 Media flagged uncertainty around whether OpenAI hashes uploaded data before processing. If the uploads aren’t hashed by the platform, those raw lists could be sitting on a server, even if briefly. OpenAI tells advertisers it deletes uploaded files within 24 hours of processing — a claim that comforts but still leaves a gap between theory and audit.
Think of a matched identifier as a fingerprint in wet cement: it proves someone passed by and can be used to find them again, but it doesn’t tell you everything about the person. The risk is not hypothetical if the deletion and hashing practices aren’t independently verifiable.
My settings said I opted out — then an ad referenced my current chat.
OpenAI says that opting out of personalized advertising removes ads based on “your past chats, activity, and preferences.” But the company also warns that ads “may still be based on your current chat.” That means the settings block some signal sources but not all.
Can I opt out of ChatGPT ads?
You can toggle personalized ads off in your ChatGPT settings. That removes ads tied to historical account signals, per OpenAI. It won’t necessarily stop ads triggered by what you type in a live session, and it won’t erase the fact that an advertiser could reach you if they matched you by email or device ID.
A campaign required a large list, but that’s not absolute protection.
Advertisers can’t target a single person; OpenAI requires audience lists to reach a minimum scale — 25,000 entries. That threshold sounds like a safety valve. But scale alone doesn’t remove the personal nature of the match, and it doesn’t answer whether the platform masks or discards the raw matching inputs quickly enough to prevent leaks.
Gizmodo asked OpenAI for details and got no immediate response.
Public reporting by 404 Media and follow-up queries from outlets such as Gizmodo have pushed the conversation into daylight. OpenAI’s documentation and public statements set out the mechanics and opt-out options, but independent verification is thin.
Advertisers are using tools and platforms that already mine identifiers — Meta’s Audience Network, Google Ads, phone number matching — and they’re bringing similar techniques into ChatGPT. The difference is that those matches are happening inside a conversational AI where you can spill real-time intent in a single thread.
Can advertisers upload my email to target me on ChatGPT?
Yes. According to reporting, advertisers can upload raw emails or hashed versions, plus phone numbers or Android advertising IDs, to create a custom audience. OpenAI’s system attempts to match those entries to ChatGPT accounts and makes matched users eligible for ads.
I recommended changes to how I use chat, and advertisers reacted within hours.
So what do you do? Use your account settings to limit personalized ads. Audit what you disclose in live chats. Consider the signals you give away when you paste emails, phone numbers, or device identifiers into prompts. And if you’re an advertiser or a security lead, ask for proof of hashing and deletion routines, not just promises.
OpenAI claims it does not reveal which advertiser matched a user, and it removes uploaded files within 24 hours. Those are welcome controls, but they’re only as strong as the company’s operational practice and independent review.
I’ve seen platforms move fast to offer new ad surfaces — from Facebook to Google to TikTok — and ChatGPT looks like the next battleground. One image is a private mailbox with a slightly ajar door: messages can get in quietly, and you won’t always know who slipped them there.
There’s no silver-bullet fix short of changing how personal data is shared and matched across services. But public pressure, audits, and clear documentation from OpenAI — plus questions from outlets like 404 Media and Gizmodo — push the product toward transparency.
Do you want the convenience of targeted recommendations in your chat, or do you want an inbox that feels truly private?