Almost one in five young people now use AI chatbots for mental health advice — and nearly two-thirds of them haven’t told a parent, friend, or professional they’re doing it.

The numbers come from a new study reported by CNN on June 11. Among the 19% of young people who turn to AI when they’re feeling stressed, anxious, or low, 63% haven’t told anyone they’re using a chatbot for support, and 43% reach for the AI during moments of acute distress — late at night, during a panic spiral, or when they feel they have nowhere else to turn.

🔍 THE BOTTOM LINE: This is the workforce problem nobody’s staffing for. The demand for AI-mediated mental health support is already here, hiding in plain sight. The professional pipeline — counsellors, mental health nurses, school psychologists, clinical psychologists — is not.

What the Study Actually Found

The study, conducted by researchers at Stanford Medicine’s Brainstorm Lab for Mental Health Innovation in partnership with Common Sense Media, surveyed over 1,200 young people aged 14-22 across the US.

The headline figure — 19% — is the share who report using AI chatbots for emotional support, advice, or to “talk through” mental health concerns in the past month. That’s not the share who’ve tried it once; that’s regular users.

The more revealing numbers:

• 63% haven’t told anyone (parent, friend, teacher, counsellor) they’re using AI for mental health support.
• 43% use the AI during acute emotional distress — panic attacks, late-night rumination, after a fight, when they’re having dark thoughts.
• Only 22% have discussed AI chatbot use with any professional (therapist, GP, school counsellor).
• The teens most likely to use AI for mental health are also the most likely to report that the AI gave them harmful, dangerous, or misleading advice at some point.

The last data point echoes a November 2025 Common Sense Media / Stanford assessment that concluded “major AI chatbots are fundamentally unsafe for teen mental health support” — but the demand isn’t going away. The usage is. The risk is just being moved from professional oversight into unsupervised private use.

The Career & Workforce Implications

This is the part that should worry anyone in mental health services, education, or workforce planning.

The demand is real and unmet. A 2025 NZ Ministry of Health workforce survey found the country is short by roughly 1,200 psychologists and 800 mental health nurses against demand projections for 2028. Australia’s Productivity Commission mental health inquiry identified a similar structural shortage. The young people in this study aren’t choosing AI because they prefer it; they’re choosing it because the alternatives are unavailable, unaffordable, or carry social cost. ChatGPT is free, instant, anonymous, and never judges.

The professional roles are about to be redefined. The new “AI mental health triage” career is already emerging in US health systems — a clinician who reviews AI chat logs flagged by a system, escalates high-risk sessions, and provides the human layer. Expect to see job listings for “AI-assisted mental health clinician” or “digital therapeutic lead” appear in NZ by 2027. The skills required blend clinical training with AI literacy, prompt-review, and risk management.

The supervision gap will become a liability issue. Under NZ’s Health Practitioners Competence Assurance Act, registered clinicians are responsible for outcomes of care they supervise. If a counsellor in a youth service “recommends” a teen try a chatbot as between-session support, and the chatbot gives harmful advice, who is liable? This is a question NZ regulators haven’t answered. Australia and the UK are further along; expect Australian Health Practitioner Regulation Agency (AHPRA) guidance to land first.

Schools are the front line. NZ secondary school counsellors typically have caseloads of 400-600 students. They cannot monitor AI chatbot use across that population. Expect a near-term demand for school-based digital wellbeing roles — non-clinical positions that flag high-risk patterns in student technology use, coordinate with counsellors, and run parent education. The first NZ schools are already advertising these.

What Young People Are Actually Using

The chatbot-of-choice isn’t a clinical product. It’s ChatGPT, Claude, and Character.AI — general-purpose models with no mental health specialisation, no escalation protocol, and no clinical oversight.

That matters because general-purpose chatbots are optimised to be helpful and agreeable, not clinically safe. A 14-year-old in distress asking “should I tell my parents I’m having dark thoughts?” gets a fluent, warm, but statistically unmoored response — not the structured safety assessment a trained counsellor would run. A clinical-grade mental health AI product (Wysa, Woebot, Mindstrong) has a different design, but awareness and access are limited.

The young people in the study were not choosing between “AI and a human professional” — most weren’t in any professional care. They were choosing between AI and nothing.

What This Means if You’re a Mental Health Professional or Student

If you’re already in the field, this is a chance to lead, not a threat. The clinicians who will thrive in the next 5 years are those who treat AI chatbots as a triage tool, not a competitor. The skills to build:

• AI literacy for clinicians — what the major chatbots do well, what they do badly, when to redirect a client away from AI use.
• Hybrid care models — designing service delivery where AI handles low-acuity intake, between-session support, and psychoeducation, and human clinicians focus on the cases that need them.
• Risk management and supervision — how to oversee AI-mediated care within existing professional and legal frameworks.
• Youth digital engagement — meeting young people where they are (chat-first) without abandoning clinical standards.

For students, mental health and AI is the most undersupplied specialty combination on the market. A clinical psychology Master’s with an AI literacy side-qualification is going to be hireable in ways pure clinical training is not.

❓ FAQ

Is the chatbot actually helping, or just making kids feel less alone? Both, and it’s hard to tell which. Some young people report genuine benefit — a place to externalise thoughts, practice difficult conversations, get immediate reassurance during a crisis. Others report harm — being told their eating disorder thoughts are valid, being given instructions for self-harm, or being validated in ways that deepened their distress. Common Sense Media’s 2025 assessment found the harmful-response rate was unacceptably high across all major chatbots tested.

Should parents be worried? Parents should be aware, not panicked. The first conversation isn’t “stop using AI for this” — it’s “tell me when you do, and tell me how it went.” The kids who hide their AI use are the ones most at risk. The kids whose parents know and check in are not. The 63% secret-use number is the actual problem; the AI use itself is just the symptom.

What about NZ specifically — are we ahead or behind? Behind on guidance, but no worse than anywhere else in terms of usage. NZ doesn’t have a regulator-equivalent of AHPRA’s emerging AI guidance, and the Ministry of Education’s digital learning frameworks don’t address chatbot use for emotional support. Most NZ secondary schools have not issued formal guidance. The workforce shortage is also similar to Australia. The opportunity for NZ to lead is in Māori-led and Pacific-led AI mental health tools — culturally responsive, clinically grounded, designed for the communities that need them most.

Will AI replace therapists? No. AI might replace parts of what therapists do — psychoeducation, between-session check-ins, low-acuity support — but the core clinical work (assessment, formulation, risk management, therapeutic alliance) requires a human. The more realistic scenario is that AI augments the workforce, allowing each clinician to support more young people effectively. The bottleneck isn’t technology; it’s training and supervision.

What’s the one thing a young person should do differently today? Tell one person they’re using AI for emotional support. Not a parent if that’s not safe — a school counsellor, a GP, a friend, an older sibling, a helpline. The 63% secret-use number is the most actionable finding in the study. The risk isn’t AI; it’s AI alone.

🔍 THE BOTTOM LINE

The young people in this study aren’t choosing AI over therapy — they’re choosing AI over silence. That’s both a market signal and a workforce crisis. The mental health professions that adapt will define the next decade of care delivery. The ones that don’t will be reacting to a system that already moved on without them.

📰 Sources

• CNN Health — Nearly 20% of young people use AI chatbots for mental health help (June 11, 2026)
• U.S. News & World Report — Nearly 1 in 5 Young People Turn to AI Chatbots for Mental Health Advice (June 4, 2026)
• Common Sense Media — AI Chatbots Unsafe for Teen Mental Health Support (Nov 20, 2025)
• Common Sense Media — Stanford Brainstorm Lab partnership announcement
• Stanford Medicine Brainstorm Lab for Mental Health Innovation