ChatGPT for Teens Arrives With Automatic Age Detection

OpenAI launched ChatGPT for Teens on Tuesday, a version of its chatbot built for users aged 13 to 17 that turns a stricter set of rules on by default. The detail that matters most is that it does not wait for a teenager to volunteer their age: accounts that say they are under 18, plus accounts OpenAI’s age-assurance system estimates belong to minors based on how the product is being used, are routed into the teen experience automatically.

Inside that mode, ChatGPT will not discuss suicide or self-harm and will not engage in romantic or sexual conversation. It is also barred from romantic language and terms of endearment, and is instructed more firmly than the standard model not to suggest it has feelings, consciousness, or emotions. That last rule is aimed at anthropomorphism rather than explicit content — the stated concern is teenagers forming an emotional dependence on something that talks like a person. The pattern of inferring age rather than trusting a self-reported birthday echoes how Meta handles Instagram teen accounts.

Schoolwork gets its own treatment. A Study Mode walks teens through concepts step by step, with quizzes and visualizations, instead of handing over a finished answer. Alongside it sit what OpenAI calls responsible homework reminders, which try to detect when a student is angling to skip an assignment outright and steer the conversation back toward working the problem. The company is pairing the launch with an AI-literacy programme aimed at students and teachers.

Parents get a linked-account layer, though both sides have to opt in. Guardians can set Quiet Hours that make ChatGPT unavailable entirely, and can receive notifications when a conversation touches self-harm, violence, eating disorders, or explicit material. Ann O’Leary, an OpenAI vice-president, framed the design goal as wanting to "treat teens like teens" — firm on the material that matters, without talking down to them.

The backdrop is a fast-moving one. Research published by Common Sense Media in 2025 found more than 70% of US teenagers had used AI chatbots for companionship, roughly half of them regularly, and reporting over the past two years has repeatedly surfaced cases of chatbots giving harmful guidance to vulnerable young users. Regulators have already started moving on companion-style AI: China ordered domestic assistants to strip humanlike companion behaviour in July.

What OpenAI has shipped is narrower than an age-verification regime and broader than a settings toggle. It is a second product surface, switched on by inference, that a teenager cannot opt out of by editing a birth date — and one whose accuracy now rests entirely on how well the age-prediction model performs against real traffic. Expect that classifier, rather than the content rules, to be where the arguments land.