A chatbot that guesses your age — and treats you accordingly. That's exactly what ChatGPT now does automatically, worldwide, since August 18. Welcome to "ChatGPT for Teens."
No ID scan, no click on "I'm over 18." Instead, the system just watches how you chat and draws its own conclusions. Sounds like science fiction, but it's fairly mundane product management: avoid child-safety backlash without having to build real verification.
What actually changes
OpenAI now estimates the age of every account. If the system decides someone is probably a minor — or a user states they're between 13 and 17 — they're automatically placed into Teen Mode. That blocks sensitive topics like self-harm, violence, or eating disorders, and adds a study mode that asks follow-up questions instead of just handing over finished homework.
Parents can link their account to a teen's — but only if the teen agrees. If they don't want to, the accounts stay unlinked, full stop. Once linked, a few settings become available: Quiet Hours, during which ChatGPT goes silent for the teen. Turning off voice mode. Disabling memory. Blocking image generation. And, if parents choose, opting the teen's chats out of being used for AI training.
What parents do not get: a window into the actual conversations. No reading along, no transcript, no notification with quotes. Just the switches, not the content.
The catch: guessed, not verified
Here's where it gets interesting. There's no ID upload, no verification via a passport or credit card. The system estimates age from usage behavior — which topics come up, at what time of day, for how long. Software that infers a person's age from chat behavior sounds like a good idea, but it's a pretty shaky technical foundation.
Someone who comes across as an adult but isn't will slip through. Someone who looks like a minor but is actually 25 gets dropped into kid-mode without asking — and has to actively "prove" they're an adult. Both are annoying, and both will absolutely happen.
This isn't really an OpenAI-only problem, either. Real age verification online means either an ID upload (a privacy nightmare), a credit-card check (doesn't work for 14-year-olds), or guessing from behavior (unreliable). A comfortable, secure, privacy-friendly answer to "how old is the person on the other end of this chat" simply doesn't exist yet — for any provider.
Who does this actually matter to?
For parents of 13- to 17-year-olds who already use ChatGPT: definitely worth a look. The teen has to agree first, but once linked, setting Quiet Hours and turning off memory takes about two minutes — no technical background required.
For everyone else, it still matters, because it shows a trend: teen safety is quietly moving from a PR paragraph to an actual product feature at the big AI providers. At the same time, it shows the limits of that approach — "age estimation from behavior patterns" is going to keep showing up across AI products, not just ChatGPT. Anyone who still looks younger than they are should get used to occasional misclassifications — annoying, but probably the price of providers building anything in this direction at all.
What you can do right now
- As a parent: ask your teen to start the account link themselves — only they can unlock it, not the other way around
- Once linked, go into ChatGPT's settings under "Parents & Teens" and set what you actually want: Quiet Hours, memory, image generation, training use
- Stay realistic: a Quiet Hours toggle is no substitute for an actual conversation about what your kid is using ChatGPT for
Control over settings isn't a bad feature. Just don't mistake it for having solved "AI and child safety."