Two reports from 404 Media describe how OpenAI's advertising system allows anyone — from large companies to essentially random individuals — to inject ads into ChatGPT sessions based on a user's phone number, email address, or unique Android advertising ID. The reports agree that while this kind of targeting is standard across the advertising industry, ChatGPT's roughly one billion monthly users may not realize that an ad they see could be there because someone uploaded their personal data to OpenAI.
The second, more detailed report adds that ChatGPT's ad targeting is also based on the prompts users type into the chatbot — prompts that users may assume are private but which, as 404 Media previously showed, may be reviewed by humans. In its own test, 404 Media ran two ads reading "ChatGPT isn't private" and targeted them to people searching for privacy-related topics. OpenAI lets advertisers provide "context hints" such as keywords, though the company says these are not exact-match rules.
Where the two sources differ is scope: the first briefly mentions unrelated research on zero-click malware delivered through online ads, while the second focuses entirely on ChatGPT's ad ecosystem, including the emergence of third-party tools like ChatGPT Ad Library, which has captured tens of thousands of ads. Both sources quote privacy researchers calling for stronger data broker laws, and both highlight that opting out of ad personalization does not stop ads or tracking entirely — as OpenAI's own setting warns, "Ads may still be based on your current chat."