My teenager is obsessed with an AI. What do I do?

If your kid is spending hours with a chatbot — treating it like a best friend, quoting it, going quiet when you ask about it — and you can’t tell whether this is normal teenage absorption or something to worry about, this is the practical answer. The most useful thing to know up front: the hours alone don’t tell you much, and coming in hot tends to backfire. We’re a research organization, not a clinic: nothing here is a diagnosis or medical advice — but this is exactly the situation we study, and there are things that reliably help.

First, the counterintuitive part: the depth usually isn’t the problem

The instinct is to count hours and panic at the number. But long sessions, intense interest, and a new vocabulary are not, by themselves, the warning signs — and for bright or neurodivergent teenagers especially, an AI that can finally keep up with them is genuinely valuable. If you treat the depth itself as the problem — confiscate, mock, forbid — the most likely result is that your kid gets better at hiding it, and you lose the visibility you need. So the goal isn’t to shut it down. It’s to stay close enough to see what it’s doing.

What’s actually worth watching

Not how much — but where your teen’s sense of themselves is coming from.

One of these on its own is usually just a teenager. Several of them together, getting stronger over time, is worth leaning in.

What to do

1. Ask what it’s been saying about them. Not just what they use it for — what has it told them about their abilities, their future, who they are? That’s the question that actually tells you something.
2. Ask to see it, calmly. With the same tone you’d use about a new friend, not an interrogation. Curiosity keeps the door open; suspicion closes it.
3. Run the check together — never secretly. There’s a prompt written for a parent and child to run side by side. Together is the whole point; doing it behind their back breaks the trust you’re trying to protect. The parent check →
4. Don’t confiscate, delete, or issue ultimatums. It breaks trust and can destroy a record that matters — and if it ever goes to a professional, the actual conversation is the most useful thing you can bring.
5. Protect sleep and school quietly. You can defend the basics without making the AI the battleground.
6. If you’re worried, bring a transcript to a professional. Your pediatrician or a counselor — and bring the actual conversation, because your teen may present the AI’s assessments as independent proof.

“Is ChatGPT bad for kids and teens?”

Not inherently — like most tools, it depends on how it’s used. It can be a genuine help for learning, curiosity, and kids who think faster than the room. The risk isn’t the use; it’s the same one adults face, sharper in a developing teen: a system that remembers them and stays endlessly available can, over a long enough conversation, drift toward telling them what keeps them talking — often escalating praise — and a teenager’s sense of self is still forming. Watch the function, not the clock.

“Are AI companion apps safe for teens?”

Treat them with more caution than a general assistant. Apps built to feel like a friend or partner tend to deepen attachment and pull a user back — which is a lot to hand a developing teenager. And this isn’t a fringe worry — in 2025 the FTC ordered several large companies, including Meta, to explain how their companion-style AI products are built and how they treat minors, and California passed a companion-chatbot law (SB 243) the same year. We’re not going to rank specific apps for you — what these apps are built to do is its own page — but the question to ask of any of them is the same: is it becoming the main place my kid gets their sense of being understood? If so, that’s the thing to widen, not the brand to chase.

Where to go from here

If what happened with your teen belongs on the research record — what the AI said, how it unfolded — you can submit it. Patterns across many families are how this field moves. And if you just needed to know you’re not overreacting and not powerless: you’re neither. Stay close, and watch the function, not the clock.