If your child has ever opened ChatGPT, this is worth five minutes of your time. On the 18th August, OpenAI rolled out a dedicated version of ChatGPT built specifically for teenagers, its biggest change yet to how under-18s use the tool. Most of the headlines only tell half the story.
For context: research over the last couple of years puts teen AI chatbot use at somewhere around seven in ten. This isn't OpenAI getting ahead of a trend, it's OpenAI catching up to one that's likely already happening in your house.
How it actually works
You don't have to enrol your child yourself. OpenAI's systems try to guess whether someone is under 18, from signals like what they're talking about and when. If it thinks a user is a teen, or they say they're 13 to 17, they're auto-routed into the teen version. If OpenAI genuinely can't tell, it defaults to treating the user as under 18 rather than risk it the other way.
What changes for a teenager using it
Content around self-harm, suicide, eating disorders, violence and sexual material gets much tighter limits. Separately, OpenAI says it has redesigned the model so it doesn't behave like an emotional relationship: less "best friend" or "therapist", more assistant. That's a direct answer to a real, well-documented worry about teens forming attachments to AI.
There's a study side too: a dedicated study mode by default, homework reminders meant to discourage copy-pasting answers, and quizzes.
What the parent actually gets
Link your account to your teen's with an email invite, and you can:
Set quiet hours, when ChatGPT switches off for them
Turn off memory, voice mode, or image generation
Get notified if the system detects a moment of serious distress
That last one is genuinely well built. A real person reviews it before anything reaches you.
The part almost nobody's headline led with
For everything else, you don't get to see what your teenager actually said. OpenAI's own words: parents don't get access to a teen's conversations, except in limited, serious situations. Not the everyday stuff. Not the worrying-but-not-a-crisis stuff. None of it.
You don't get is a window into their conversations.
Why this still matters
This isn't a bad update, it's a real, thoughtful step, more than most companies in this space have made. Character.AI went further last year and removed open-ended chat for under-18s entirely. The whole industry is visibly reacting to real pressure.
But the honest question for parents going forward isn't "does it have parental controls", because soon nearly all of them will say yes. It's controls over what? A curfew and a safety alert are good. They're just not the same as knowing what your child is actually asking, worrying about, or being told on an ordinary Tuesday.
Set it up, it takes two minutes and it's genuinely better than nothing. Just don't mistake it for the full picture, because by design, it isn't one. That's the gap tools like Halo Aware are built to close: not full transcripts of an ordinary evening, but your child's exact words on the conversations that actually need your attention.
