Our approach to age prediction

OpenAI wants ChatGPT to know how old you are, even if you never tell it. The company is deploying an age-prediction system that scans account signals and usage patterns to guess whether someone is above or below 18. Accounts that get flagged as likely teens will automatically receive a more restricted version of the chatbot, with tighter content safeguards baked in.

This isn't a one-time check at signup. OpenAI describes it as an ongoing process, refined as more data comes in and the model gets better at telling adults from minors. That matters because the whole system rests on probability, not proof. Nobody is scanning IDs here. The company is instead betting that behavioral and contextual clues, things like how someone phrases questions or what topics they gravitate toward, can approximate an age with enough confidence to act on it.

The timing lines up with a broader reckoning across the industry over how AI chatbots interact with younger users. Regulators and parents alike have pushed companies to prove they're not handing unrestricted access to systems capable of discussing sensitive or adult topics. OpenAI's answer is prediction plus friction: guess age first, then adjust what the product will and won't do, rather than relying purely on self-reported birthdates that anyone can fudge in ten seconds.

The tricky part is what happens when the guess is wrong. An adult misclassified as a teen suddenly finds their chatbot behaving like it's talking to a 15-year-old, filtered and cautious in ways that might feel patronizing or just useless for their actual needs. OpenAI hasn't detailed an appeals process or how often these predictions miss the mark, only that accuracy will improve with time. For now, this is a rollout built on inference, and the company is asking users to trust the system while it learns.