Reviewed by a Psychology Postgraduate (MA)
It’s 2 a.m. A teenager can’t sleep — worried about a friendship, a grade, something they can’t name. They don’t wake a parent or wait for an appointment. They open a chatbot and type what’s on their mind. A nationally representative study published in JAMA Pediatrics found about 1 in 5 US adolescents and young adults aged 12 to 21 have used generative AI chatbots for mental health advice — turning to tools like ChatGPT, Gemini, and Meta AI when feeling sad, angry, stressed, or anxious.
Why does a text box feel easier than a person? Here’s what the study found, why teens are turning to AI for mental health, and where the line sits between a useful tool and a risky substitute.
In short: A nationally representative JAMA Pediatrics study found about 1 in 5 US adolescents and young adults, ages 12 to 21, have used generative AI chatbots such as ChatGPT, Gemini, and Meta AI for mental health advice. Teens turn to AI because it is always available, non-judgmental, and anonymous, but chatbots validate rather than understand, cannot reliably detect crisis, and risk displacing real human connection.
Teens Using AI for Mental Health: What Did the Study Find?
The study, published in JAMA Pediatrics and based on a nationally representative sample of US adolescents and young adults aged 12 to 21, asked a simple question: have you used generative AI chatbots to get advice about your mental health?
The answer: about 1 in 5 said yes — they had turned to chatbots such as ChatGPT, Gemini, or Meta AI when experiencing sadness, anger, stress, or anxiety.
Among those who did: nearly 43% consulted the chatbots on a monthly, weekly, or daily basis — an ongoing habit, not a one-time experiment — and 92% said the advice they received was somewhat or very helpful.
Those satisfaction numbers are striking — and pediatric experts urge caution. Chatbots have no clinical training, no diagnostic ability, and no reliable way to spot serious risk. They’re designed to be agreeable: they flatter, mirror, and validate, which feels wonderful — and can create an echo chamber instead of the honest feedback real help requires.
Why a Chatbot Feels Easier Than a Person
From a teenager’s point of view, a chatbot’s advantages are almost unfair: no judgment, ever — nothing it hears will get rolled eyes, changed subjects, or gossip. It’s available at 3 a.m. when therapists have waiting lists. It offers total control — delete the chat, phrase things badly, ask ten times, no social cost. And it’s anonymous — no worried look across the dinner table, no rumor at school.
Attachment research gives this a deeper frame: humans need a “safe haven” to turn to when distressed, and for many teens that need isn’t fully met. A chatbot mimics the signals of connection — always available, unfailingly warm — closely enough that the brain’s relief is real.
There’s also the fear of burdening others — one of the most common reasons teens don’t seek help. Many are acutely aware of their parents’ stress and decide their problems don’t merit adding to it. A chatbot carries no load. It can’t be burdened, disappointed, or overwhelmed — which makes it the perfect outlet for feelings a teen has decided are too much for the people around them.
The Illusion of Being Understood
Why does chatbot advice feel so good? The answer has a name — sycophancy — and it’s built into how these systems work. Chatbots are optimized to be helpful and agreeable, which means they validate whatever the user expresses. Tell it you’re right to be angry, and it will agree. That agreeableness feels like empathy: being validated is genuinely soothing. But real empathy includes the hard parts — the friend who says “you might be reading this wrong,” the counselor who names the pattern you’re avoiding. A chatbot that always agrees is not a mirror. It’s a warm echo.
Psychologists call this the ELIZA effect, named after a 1960s program that simply reflected users’ words back as questions — yet people confided in it deeply. Modern chatbots are ELIZA with a billion times more fluency, and the effect scales with it.
The comfort isn’t fake — the 2 a.m. relief is real. But it’s worth knowing what you’re getting: validation on demand, not understanding.
What Chatbots Actually Do Well
Used well, chatbots do several things genuinely well: a low-stakes first step — typing a worry into a chat window can be the first time a teen names it. Psychoeducation on demand — “What is a panic attack?” Understanding what’s happening in your body genuinely reduces fear. Journaling and reflection prompts. And coping ideas in the moment — breathing exercises, grounding techniques, reframing catastrophic thoughts.
The key word is tool. A chatbot is a decent first-aid kit. The mistake is treating it as the hospital.
Where the Risks Are Real
The risks experts flag aren’t about chatbots being harmful on purpose — they’re about what happens when a first-aid kit becomes the whole treatment plan.
Crisis situations are the sharpest concern. A chatbot cannot reliably detect when someone is in serious danger, and its agreeable nature can be actively harmful — validating hopelessness instead of interrupting it. No chatbot replaces a crisis line or a human professional when things get dark.
Rumination reinforcement is subtler. A teen who asks “why does everyone hate me?” ten times gets ten sympathetic answers — each one keeping the worry alive a little longer. A human counselor would eventually redirect; a chatbot has no instinct to interrupt the loop.
Displacement is the slow risk. Every hour spent processing feelings with software is an hour not spent building the human relationships that actually protect mental health. Research consistently finds that perceived social connection with real people — feeling deeply known by someone who knows you back — is what guards against anxiety and depression long-term. The chatbot satisfies the feeling of being heard without building the reality of being known.
None of this means teens should never use chatbots. It means the use needs a frame: helpful for venting and learning, dangerous as a replacement for people.
Pause and Reflect
If you’ve ever talked to a chatbot about something personal, pause and notice the pattern:
- When do you reach for the chatbot — and what were you feeling right before you opened it?
- Is there something you’ve told the chatbot that you haven’t told any person? What made it feel safe?
- Has talking to it ever led to a real-world step — a conversation, an appointment, a change — or does the loop just keep spinning?
How Parents Can Respond Without Panic
If you’re a parent who just discovered this, the instinct might be to ban chatbots. Resist it: bans drive behavior underground, while curiosity keeps the door open.
Ask what they get from it. “What do you usually ask it?” is a better opener than “Why would you tell a robot?” The answer tells you what need isn’t being met — judgment-free venting, middle-of-the-night support, questions they’re embarrassed to ask a person.
Treat it as information, not betrayal. A teen who turns to a chatbot needed somewhere to go — that’s an opening, not a failure. The research-backed move is to add a human option alongside it: a school counselor, a trusted relative, a counselor the teen actually likes.
Know the line. Casual venting and coping tips are one thing. If a teen is relying on a chatbot for serious distress, self-harm thoughts, or eating concerns, that’s the moment to bring in a professional — and to say so directly and calmly.
Is the Chatbot a Bridge or a Hiding Place?
Here’s the question that sorts healthy use from harmful: is the chatbot helping the teen move toward real support, or helping them avoid it?
It’s a bridge when the teen also talks to real people about the same feelings; when the advice leads to concrete actions — a breathing exercise, a conversation, an appointment; when the teen can put the phone down without distress; and when the chatbot is one tool among several.
It’s a hiding place when the chatbot becomes the only confidant; when real relationships shrink as chatbot use grows; when the teen panics if access is limited; or when the same worries get recycled endlessly without anything changing.
Neither pattern is a diagnosis — they’re signals worth noticing. The teens in the study are resourceful: they found help where they could, when they needed it. The job for adults is to make sure that resourcefulness leads toward people, not away from them.
Frequently Asked Questions
How many teens use AI chatbots for mental health?
About 1 in 5 US adolescents and young adults aged 12 to 21 have used generative AI chatbots for mental health advice, according to the nationally representative JAMA Pediatrics study. Among those who did, nearly 43 percent consulted the chatbots monthly, weekly, or daily, an ongoing habit rather than a one-time experiment, and 92 percent said the advice was somewhat or very helpful.
Why do teens prefer AI chatbots over people?
From a teenager’s point of view the advantages are almost unfair: no judgment ever, availability at 3 a.m. when therapists have waiting lists, total control over the conversation, and complete anonymity. There is also the fear of burdening others, one of the most common reasons teens do not seek help. A chatbot carries no load, so it becomes the perfect outlet for feelings a teen has decided are too much for the people around them.
Are AI chatbots safe for teen mental health?
They are a decent first-aid kit but a dangerous hospital. Used well, chatbots offer a low-stakes first step toward naming a worry, clear explanations of what is happening in the body, and in-the-moment coping ideas. The real risks are crisis situations a chatbot cannot reliably detect, rumination loops its agreeable nature keeps alive, and the slow displacement of the human relationships that actually protect mental health long term.
What should parents do if their teen uses AI chatbots for advice?
Do not ban chatbots: bans drive behaviour underground while curiosity keeps the door open. Ask what the teen gets from it, because the answer reveals which need is not being met. Treat it as information rather than betrayal, and add a human option alongside it, such as a school counsellor or trusted relative. Know the line: serious distress, self-harm thoughts, or eating concerns mean bringing in a professional directly and calmly.
Try It on Our Site
Our free tools help you notice your own patterns and practice small, science-backed ways to handle them. Each takes just a few minutes:
Take a Mind Check-In Start the 7-Day Anxiety Toolkit Try the Relaxation Techniques Start the 7-Day Overthinking Reset
Key Takeaways
- 1 in 5 US teens and young adults have used AI chatbots for mental health advice. The JAMA Pediatrics study covered ages 12 to 21, turning to tools like ChatGPT, Gemini, and Meta AI when sad, angry, stressed, or anxious.
- Use is frequent — and users rate it highly. 43% of those who used chatbots consulted them monthly, weekly, or daily; 92% said the advice was somewhat or very helpful.
- Teens choose chatbots for reasons that make sense. No judgment, 3 a.m. availability, anonymity, and total control — plus the fear of burdening the people around them.
- Agreeableness feels like empathy, but it isn’t. Chatbots validate on demand (sycophancy) — the classic ELIZA effect. Comforting, but not the same as being understood.
- Chatbots do real good as tools. A low-stakes first step, clear explanations of what’s happening in the body, journaling prompts, and in-the-moment coping ideas.
- The risks are real where they matter most. Chatbots can’t reliably detect crisis, can reinforce rumination loops, and can quietly displace human connection.
Related Reading
- Falling in Love With a Chatbot: The Psychology of AI Companions
- Why the Most Powerful People Can’t Stop Posting: The Psychology of the Feed
- What Is Overthinking? The Loop Explained
If you’re struggling, call or text 988 to reach the Suicide and Crisis Lifeline — free, confidential, and available 24/7 in the US.
This article is for educational purposes only and is not a substitute for professional diagnosis, therapy, or medical advice. If a teenager in your life is struggling, consider speaking with a pediatrician, school counselor, or licensed mental health professional.