ChatGPT ad targeting uses prompts, email and phone
Advertisers running ads inside ChatGPT can upload customer lists — emails, phone numbers, their SHA-256-hashed versions, or Google Advertising IDs — to match them to real ChatGPT accounts, per a 404 Media investigation. Lists of 25,000 or more matched users can be used to actively target people; smaller lists can only exclude them. Separately, advertisers can supply “context hints” — topics or keywords — that OpenAI says guide ad matching without being exact-match rules, similar to looser versions of Google’s keyword ads; 404 Media’s own test campaigns used this to target users based on words in their live conversations. Turning off ad personalization stops matching to past chats or uploaded contact lists, but OpenAI can still show ads based on the content of a user’s current, active conversation, according to the investigation. OpenAI representatives told 404 Media that uploaded customer records are deleted within 24 hours of processing.
Why it matters: OpenAI's ad targeting now runs on two separate tracks — uploaded identity data and live conversation content — and the setting most users would assume turns targeting off only covers one of them.
Via 404 Media ↗
Posted to the wire October 17, 2026. Edited by Joe Balewski.