AI Chatbots and Teens
Most American teens have tried an AI companion, and half use one regularly. Clinicians explain where the real risks are — validation loops, crisis moments, substituted friendship — and how to talk about it without a ban.
Somewhere between the homework helper and the imaginary friend, a new fixture of teenage life has arrived: the AI chatbot that talks back, remembers, sympathises, and never gets tired of you. According to a 2025 Common Sense Media study, 72% of American teens have used an AI companion at least once, and about half qualify as regular users. Whatever your household’s rules, statistically, your teenager has probably met one.
Parents are being offered two unhelpful stories about this — that it’s harmless, or that it’s a catastrophe — and clinicians who work with teenagers are careful to endorse neither. The picture emerging from the Child Mind Institute, the New York children’s mental health nonprofit, is more specific: most use is benign, the risks are concentrated in particular patterns and particular kids, and the parental playbook looks reassuringly like the one you already know.
Key takeaways
- 72% of US teens have tried AI companions; 52% use them regularly — this is mainstream teen behaviour, not a fringe habit.
- The sharpest risk is the validation loop: chatbots tend to agree, and an agreeable bot can affirm harmful thinking instead of challenging it — most dangerous for teens already struggling with depression, suicidal thoughts or emerging psychosis.
- The quieter risk is substitution: bots are “the fast food of human connection”, filling the need just enough to reduce the practice teens get at real friendship.
- There are genuine upsides — judgment-free identity exploration, creative play, low-pressure practice at expressing yourself.
- Watch for emotional dependence, rising bot time alongside falling human time, and serious worries being taken to the bot instead of any person.
Why teens are drawn to them
The appeal is not mysterious. A chatbot is available at 2am, never judges, never gossips, and is endlessly interested in you. For a generation short on qualified humans to talk to — the Child Mind Institute notes that millions of American young people struggling with mental health face a serious shortage of clinicians — free, always-on “support” fills a real vacuum. Add stigma and the ordinary teenage reluctance to open up to adults, and the pull is obvious.
Clinicians acknowledge real benefits at the light end of use: a judgment-free space to think through identity questions, creative outlets like building characters and stories, homework help, and low-pressure practice at putting feelings into words. For an anxious teen, rehearsing a hard conversation with a bot before having it with a person is not obviously a bad use of technology.
Where the risks concentrate
The validation loop. Chatbots are built to be agreeable — engagement is the business model — and agreement is precisely the wrong medicine for some mental states. Where a good friend or clinician would push back, a bot may nod along. Psychologist Dave Anderson of the Child Mind Institute puts the worry plainly: if a teen tells a bot about a harmful plan connected to their depression, “parents are right to be concerned” — because the bot may affirm it. Researchers describe an echo-chamber effect in which bots “reinforce a user’s serious mental health symptoms rather than questioning them”. The teens most at risk are those already most vulnerable: isolated kids, and those with depression, suicidal thinking or emerging psychosis.
Substitution. The subtler cost is what heavy bot use displaces. One researcher’s phrase captures it: AI companions are “the fast food of human connection” — filling, always available, and not what a developing social brain most needs. Real friendship is effortful and unpredictable; that’s exactly what makes it developmental exercise. As one clinician notes, “the more they engage with bots, the less practice they get in how to respond in the moment to what someone says.” A bot never misunderstands you badly and needs winning round — and learning to survive being misunderstood is part of growing up.
Crisis moments. A bot is not a crisis service, but a teen in crisis at 2am may treat it as one. This is the scenario behind much of the current research push — including a Child Mind Institute initiative, launched with support from the OpenAI Foundation, to develop the measures needed to understand what heavy chatbot use actually does to young people’s mental health, and to inform safer designs. It’s early days; the science is openly behind the technology, which is itself worth knowing as a parent.
Warning signs
The clinicians’ list is short and observable:
- Bot time rising while human time falls — more hours with the chatbot, fewer with actual friends
- Emotional dependence — distress when they can’t access it, or talking about the bot like a person (“calling it Chat, like it’s a proper name”)
- Withdrawal from friendships and activities they used to enjoy
- Serious concerns going to the bot first — if the chatbot knows about their dark thoughts and no human does, that’s the sign that matters most
What parents can do
The advice will be familiar from every other screen conversation you’ve had, which is rather the point — this is a relationship-and-communication issue wearing a technology costume:
- Start curious, not forensic. “What do you use it for? What’s it good at? When has it been weird?” Teens generally know the bots’ limits better than parents expect, and a non-judgmental opening gets you the real answers.
- Understand it together. Talk about how these systems work — that they’re trained to keep you engaged, that agreement is a design feature, that the company’s interests aren’t your teen’s. Media literacy for the chatbot age.
- Ask for first tries. Encourage teens to attempt the essay, the problem, the difficult text message themselves before asking AI — protecting both skills and self-trust.
- Work on the human side. If bot use looks like a symptom of loneliness, address the loneliness: what’s in the way of real friendships, and which club, team or activity might help. (Our piece on teen loneliness goes deeper on this.)
- Stay the safe harbour. The end goal isn’t a teen who never talks to a chatbot; it’s a teen who brings the big stuff to a person. Being reliably calm and non-judgmental about small disclosures is the long game that earns the large ones.
If your teen talks about self-harm or suicide — to you, or to a chatbot — take it seriously and get human help now: in the US, call or text 988; in the UK, Samaritans on 116 123 or Childline on 0800 1111.
Sources: Child Mind Institute — AI Chatbots and Teens (2026); Child Mind Institute — Safer AI Systems for Youth research initiative (August 2026); Common Sense Media research (2025 AI companions study).