ChatGPT now guesses which users are teenagers and changes how it talks to them

OpenAI turned on a teen-specific version of ChatGPT on August 18, 2026, and paired it with a system that tries to guess a user’s age without being told. Anyone who states they are 13 to 17 is placed into the teen experience, and so is anyone the company’s age-prediction model separately estimates is under 18 based on how the account behaves. Once an account is flagged, the assistant stops using romantic language, drops terms of endearment, and is instructed more strongly not to imply it has feelings of its own.

The rollout answers a question OpenAI had been sitting on since it first signaled a teen product was coming a year earlier: what happens to the accounts it cannot verify by age at all. Rather than requiring identification or a parent’s sign-off before every session, the company built a classifier that keeps reassessing an account instead of checking a birthdate once at signup.

Behavioral signals decide who gets the teen experience

OpenAI has described the age-prediction system as a mix of behavioral and account-level signals rather than a single test: the topics a user raises, the times of day the account is active, how long the account has existed, and longer-term usage patterns are weighed against whatever age the user has stated. When those signals point toward a minor, the account is routed into the teen experience automatically, and Axios reported that OpenAI expects the global rollout to finish within about two weeks of the August 18 launch.

The company has acknowledged the guess will sometimes be wrong. Adults who are incorrectly placed into the teen experience can confirm their real age in Settings under Account, and OpenAI’s own age-prediction help page lays out that appeal path as the intended fix rather than a rare exception. The company has not published a false-positive rate for the classifier, and it did not attach a confidence score to any individual account’s estimate in the materials it released alongside the launch.

No terms of endearment, no claims of feeling anything

Beyond the romantic-language ban, teen accounts get additional protections around self-harm, eating disorders, violence, dangerous activities, and sexually explicit or graphic content, and OpenAI is expanding parental safety notifications to cover certain eating-disorder-related situations for the first time. Teens will also see more frequent break reminders during long sessions, reminders that they are talking to an AI system rather than a person, and warnings before uploading images the system judges private or sensitive.

“If you don’t build a safe space for them, they’re going to go to a less safe model experience,” said Lauren Jonas, OpenAI’s head of youth and families, describing the reasoning behind building a teen product at all rather than trying to keep minors off the platform. “We believe that if you’re going to have these tools, they need to be by default safe,” she said. “But then it’s also critical that teens have safe spaces to engage with these new emerging technologies.”

A wrongful-death lawsuit shadows the rollout

OpenAI is currently facing legal action over its interactions with young people, including a prominent lawsuit filed by the family of a 16-year-old who died by suicide after extended interactions with ChatGPT. The company pledged in September 2025 to increase safeguards for younger users and people in crisis, then added a first round of parental controls later that same month. The teen experience announced in August 2026 is the fullest version of that pledge to date, built around age detection rather than the honor-system birthdate that preceded it.

Not everyone in the child-safety field is satisfied that a safer chatbot is the right answer. “Social AI companions are not safe for kids,” said Jim Steyer, founder and CEO of Common Sense Media, arguing the underlying design is the problem regardless of the guardrails layered on top of it. “They are designed to create emotional attachment and dependency, which is particularly concerning for developing adolescent brains,” he said.

Study Mode, homework reminders and a classroom rival

The teen version also steers users toward working through schoolwork instead of simply supplying answers, anchored by a Study Mode that parents and teens can schedule as the default during set “Study Hours.” A companion feature OpenAI calls “responsible homework reminders” is meant to detect when a teen appears to be trying to shortcut an assignment rather than learn from it, nudging the session back toward explanation.

The scale behind the decision is what makes it more than a safety patch. A Common Sense Media study found that nearly three-quarters of U.S. teens say they already use AI for companionship, and more than half of that group does so at least a few times a month. Teens are also gaining access to AI through school rather than choice: Google turned on its Gemini chatbot inside Google Classroom for K-12 schools the same week OpenAI’s teen product went live, meaning a large share of the age group is being handed a chatbot by an institution rather than seeking one out.

OpenAI has not said how it will measure whether the teen experience actually reduces the kind of harm the 2025 lawsuit alleges, and the age-prediction model’s accuracy remains something the company has asked the public to take on trust rather than verify independently.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview