A recurring claim has surfaced among ChatGPT users: after discussing specific topics in private conversations, they see eerily relevant ads on Facebook shortly after. The reports describe scenarios where niche subjects discussed only with ChatGPT - not searched on Google, not mentioned on social media - appear as targeted ads within hours.
There is no confirmed evidence that OpenAI shares conversation data with Meta or any advertising platform. OpenAI's privacy policy states that it does not sell personal data, and Meta's ad targeting relies on its own extensive data collection infrastructure, not third-party AI chat logs.
So what's likely happening? The most probable explanation is confirmation bias combined with how broad modern ad targeting actually is. Facebook tracks browsing activity across millions of sites through its pixel tracking code, monitors in-app behavior, and purchases data from brokers. The topics you think about enough to discuss with ChatGPT are usually topics you've already signaled interest in through dozens of other digital behaviors - browsing patterns, app usage, purchase history, location data - that feed ad algorithms independently.
That said, the concern isn't baseless from a structural standpoint. ChatGPT conversations contain some of the most candid text people produce digitally. Users share health worries, financial situations, relationship problems, and business plans. Even without any data sharing today, the question of how AI companies will handle this data long-term is worth paying attention to. OpenAI already uses conversations to train models by default unless you opt out in settings, and their enterprise and API tiers offer stronger data isolation.
If ChatGPT conversation privacy matters to you, practical steps exist right now: disable model training on your data in Settings > Data Controls, use the API instead of the consumer product for sensitive topics, or switch to a provider with stronger privacy defaults. Don't rely on coincidental ad timing as evidence of a breach - but don't assume your AI conversations are as private as you think, either.