In 2025, something crossed over quietly: among young Americans, getting mental health advice from an AI chatbot became as common as getting it from a licensed professional. That sentence deserves a page of careful, verified numbers, so here it is, with the study methodology printed up front, because on this topic the details matter more than anywhere else on this site.
The short answer: it depends who you ask about. Among US adolescents and young adults (12–21), 19.2% have used AI chatbots for mental health advice, about 8.2 million young people, statistically level with the 19.8% receiving professional counseling. Among adults, 1–3% actively use AI for mental health, 16% sought mental health information from a chatbot in the past year (28% under 30), and among AI users who report mental health challenges, 48.7% have used a major chatbot for support.
Every figure below names its study, its population, and its definition. On a topic this sensitive, I’ve also flagged which findings come from large peer-reviewed studies and which come from smaller industry surveys. Sources at the bottom.
Table of contents · 9 sections
- AI and Mental Health: Editor’s Choice
- The primary statistic, with its methodology
- Who uses AI for mental health, and how often
- Why they use it, and on which platforms
- Does it help? Users vs experts
- Growth, secrecy, and the regulatory response
- Chatbots vs companions: the distinction that matters
- Conclusion
- FAQs
AI and Mental Health: Editor’s Choice
- →19.2% of US adolescents and young adults have used AI chatbots for mental health advice, level with the 19.8% receiving professional counseling. (RAND / JAMA Pediatrics)
- →Usage jumped from 13.1% to 19.2% in a single year, growth of more than 40%.
- →63% of young users have told no one, not parents, not doctors, not friends.
- →91.7% of young users found the advice helpful, while 94% of psychologists say chatbots can’t treat with appropriate nuance. Both stats are real.
- →In independent testing, popular AI models responded inappropriately to mental health symptoms 20%+ of the time. (Stanford)
- →The demand engine: 137 million Americans live in mental health professional shortage areas, and nearly half of adults with mental illness got no treatment last year.
- →The first randomized trial of a purpose-built therapy chatbot cut depression symptoms by 51%. (Dartmouth / NEJM AI)
The primary statistic, with its methodology
The headline number comes from the best study yet run on this question, so here is exactly what it measured:
1. 19.2% of US adolescents and young adults have used AI chatbots for mental health advice.
The study card: population: US residents aged 12–21. Definition: used AI chatbots (ChatGPT, Gemini, Character.AI, Meta AI) “for advice or help when feeling sad, angry, nervous, or stressed.” Sample: 1,009 respondents, weighted to represent 42.8 million young Americans. Data year: 2025, published June 2026. That 19.2% represents an estimated 8.2 million young people.
2. Among the general adult population, active use runs about 1–3%, with over 10% expressing interest.
The strictest adult ruler: regularly and deliberately using AI for mental health, right now. Small, but the interest figure and the next two stats show the pipeline behind it.
3. 16% of US adults sought mental health information or advice from an AI chatbot in the past year, rising to 28% of adults under 30.
The broader ruler: any mental-health-related use in twelve months, part of the 32% of adults who used AI for health information generally. Between the 1–3% active core and this 16% annual reach lies the usual adoption story: occasional use first, habits later.
4. Among AI users who report mental health challenges, 48.7% have used a major chatbot for support.
The population that matters most: people already using LLMs who also report mental health struggles. Nearly half have pointed the chatbot at those struggles. Where the need and the tool coexist, usage is already a coin flip.
Who uses AI for mental health, and how often
The RAND study breaks its 19.2% down, and two of its findings run against expectations.
5. Young women are 2.1× more likely to use AI for mental health advice, and 18–21-year-olds are 3.7× more likely than 12–14-year-olds.
The demographic shape: older teens and young adults, skewing female, the same profile that dominates therapy-seeking generally. This isn’t a fringe cohort adopting a fringe tool; it’s the help-seeking population extending its help-seeking to a new channel.
6. The counterintuitive finding: youth who recently discussed mental health with a physician were MORE likely to use chatbots, not less.
1.9× more likely. The substitution story, kids ditching doctors for bots, isn’t what the data shows. For most young users this is a supplement layered on top of care, which is also what psychologists report from the other side (stat 18).
7. Among young users, 42.8% seek AI mental health advice at least monthly.
Not a one-time experiment. For over four in ten of them, this is a recurring habit, and one industry survey of adult users puts the split at roughly 22% daily and 38% weekly, figures I’d treat as directional rather than definitive given the smaller sample.
8. Honest gaps: older adults, college students specifically, and urban vs rural use haven’t been credibly measured yet.
The research wave is young. If you see precise percentages for those groups on other stats pages, check the source before trusting them; as of this writing, they don’t trace to peer-reviewed data. This page will be updated as the studies land.
Why they use it, and on which platforms
Two forces explain nearly all of it: what the moments feel like, and what the alternatives cost.
9. The measured use cases: feeling sad, angry, nervous, or stressed, plus fear of judgment as a documented driver.
The RAND definition captures everyday emotional weather, not clinical crises. Smaller surveys add the why-a-bot layer: roughly 1 in 3 users cite fear of being judged by a human. A chatbot has no face to disappoint, which for some people is the entire product.
10. The access gap underneath: nearly half of US adults with mental illness received no treatment last year, and 137 million Americans (40%) live in professional shortage areas.
The single most important context on this page. Before judging anyone for consulting a chatbot, note that for a huge share of the country, the realistic alternative wasn’t a therapist. It was nothing.
11. The economics: therapy commonly runs $100–200 per session; the chatbot is free, instant, and awake at 3 a.m.
Waitlists for therapists stretch weeks to months; the loneliest hours, as my loneliness statistics showed, don’t keep office hours. Cost, wait time, and availability explain the adoption curve better than any enthusiasm for AI does.
12. The platforms in the research: ChatGPT, Gemini, Character.AI, and Meta AI, mostly general-purpose tools.
Note what tops the list: not purpose-built therapy apps but everyday chatbots, plus Character.AI, where a single user-created “Psychologist” bot has received tens of millions of messages. How many users that platform really has is its own statistical rabbit hole, and the next page in this series.
13. The scale of acute need: OpenAI disclosed that over 1 million people per week show signs of severe crisis in ChatGPT conversations.
Stated soberly, because it should be: general-purpose chatbots are now, de facto, a front line for people in serious distress. That reality, not hype, is why safeguards, crisis routing, and the regulation in section 6 exist, and why every platform’s handling of these moments deserves scrutiny. If that describes you, please reach a human: call or text 988 in the US.
Does it help? Users vs experts
Here’s the strangest chart on this page: the people using it and the people trained to provide it have nearly opposite verdicts, and both are backed by data.
14. 91.7% of young users rated the AI’s mental health advice as somewhat or very helpful.
Near-unanimous satisfaction, consistent with the loneliness research finding that AI conversation genuinely delivers the feeling of being heard. Satisfaction is real data. It just isn’t the same thing as clinical safety, hence the rest of this section.
15. The first randomized controlled trial of a purpose-built therapy chatbot cut depression symptoms by 51% and anxiety by 31%.
The strongest clinical evidence to date, and the crucial nuance: Therabot was built by clinicians, trained on evidence-based therapy, and run with human oversight. The trial validates purpose-built AI therapy, not the general chatbots doing most of today’s volume.
16. In independent testing, popular AI models responded inappropriately to mental health symptoms at least 20% of the time, and 41% of users report receiving misleading advice.
The other side of the 91.7%. One inappropriate response in five is a poor rate for a barista and an alarming one for a de facto counselor, and users themselves notice the errors. Feeling helped and being helped can diverge, which is precisely the experts’ concern.
17. 94% of psychologists say chatbots cannot treat conditions with the appropriate level of nuance.
From a survey of 1,200+ licensed US psychologists. Near-consensus skepticism about AI as treatment, though notably the same survey shows the profession engaging with the reality rather than dismissing it, see the next stat.
18. 35% of psychologists report having patients who use AI as an “additional mental health professional.”
More than a third of therapy rooms now contain three parties: patient, clinician, and a chatbot at home. The clinical world’s emerging position is supplement-with-oversight, never substitute, which matches what the RAND physician finding suggested from the patient side.
Growth, secrecy, and the regulatory response
Three numbers define where this is heading: one about speed, one about silence, and one about the law.
19. Youth usage grew from 13.1% to 19.2% in one year, pulling statistically level with professional counseling (19.8%).
More than 40% growth in twelve months, against a counseling rate that barely moves because it’s constrained by workforce, cost, and geography. The chatbot has none of those constraints, which is why the crossover was less a surprise than a schedule.
20. 63% of young users have told no one they use AI for mental health advice.
Not parents, not physicians, not friends. Two-thirds of this entire phenomenon is happening in private, which means most parents and clinicians are advising young people while blind to one of their main sources of advice. The researchers’ own recommendation: ask. It’s the cheapest safeguard that exists.
21. The law arrived: Illinois became the first state to ban AI-delivered therapy in 2025, with Nevada and Utah restricting it and six more states advancing bills.
The emerging legal line mirrors the clinical one: AI out of the therapist’s chair, AI permitted for administrative and wellness support. Expect the map to keep filling in through 2026 and 2027.
22. The market grows regardless: AI mental health is projected from $2.1 billion in 2026 to $9.1 billion by 2033.
A 4× expansion right through the regulatory turbulence, because the demand documented in section 3 isn’t going anywhere. The open question isn’t whether AI stays in mental health; it’s whether the purpose-built, clinically supervised version wins out over the general chatbots currently doing the volume.
Chatbots vs companions: the distinction that matters
One distinction the statistics above blur constantly, and my corner of this industry lives exactly on it.
23. Three different things get counted as “AI mental health use”: general chatbots, purpose-built therapy tools, and AI companions.
General chatbots (ChatGPT, Gemini) handle most advice volume; purpose-built tools (the Therabot class) carry the clinical evidence; and companion apps are where emotional support becomes an ongoing relationship, 67% of their users cite loneliness as their motive, and peer-reviewed work finds they reduce it about as much as human interaction. Same technology, three different risk profiles, and lumping them together is how both the hype and the panic get manufactured. The companion side’s full numbers are in my AI companion statistics.
The emotional-support side of AI is the market I test every day.
Companion apps sit at the exact intersection this page describes: always available, judgment-free, and increasingly where people take their feelings, which is why I pay to test them myself and grade them on safety and privacy, not just features. The market data is in my AI girlfriend statistics, and the hands-on comparison is on my homepage.
Read the AI girlfriend statistics →Conclusion
So, what percentage of people use AI chatbots for mental health advice? Nearly 1 in 5 young Americans, level with professional counseling and growing 40% a year; 16% of adults in the past year; nearly half of AI users who carry mental health challenges. They find it helpful at rates the experts find alarming, and the law has started drawing lines through the middle of it.
My read as a data researcher: the crossover stat is the quiet mental-health shift of the decade, but the number I’d watch is the 63% secrecy figure, because everything risky about this trend gets riskier in the dark. And the honest framing of the 19.2% is that it isn’t a verdict on AI at all. It’s a verdict on access: 137 million Americans live where therapists are scarce, and the chatbot was simply the first thing that always picked up. If your own struggles run deeper than a rough evening, please let a human in on it too, a professional if you can, and in a crisis, 988.
FAQs
What percentage of people use AI chatbots for mental health advice?
Among US youth aged 12–21, 19.2% have used AI chatbots for mental health advice (RAND/JAMA Pediatrics, 2025 data), about 8.2 million young people. Among adults, roughly 1–3% are active regular users, while 16% sought mental health information from a chatbot in the past year, rising to 28% of adults under 30 (KFF).
Do AI chatbots actually help with mental health?
The evidence runs both directions. 91.7% of young users rate the advice as helpful, and the first randomized trial of a clinician-built therapy chatbot cut depression symptoms by 51%. But independent testing found popular general-purpose models responding inappropriately to mental health symptoms over 20% of the time, and 94% of psychologists say chatbots can’t treat with appropriate nuance. Purpose-built, supervised tools show real promise; general chatbots are less reliable than they feel.
Is it safe to use AI for mental health advice?
For everyday stress and processing feelings, millions use it without incident, and researchers’ main advice is transparency: tell someone you’re using it, since 63% of young users currently don’t. For diagnosed conditions, crises, or thoughts of self-harm, the answer is no, AI is not a safe substitute: reach a professional, a trusted person, or in the US, call or text 988, which is free and available around the clock.
Why do people use AI instead of a therapist?
Mostly because the therapist wasn’t available: nearly half of US adults with mental illness received no treatment last year, 137 million Americans live in provider shortage areas, sessions cost $100–200, and waitlists run weeks. Add zero judgment and 3 a.m. availability, and the adoption curve explains itself. Notably, research shows most young users treat AI as a supplement to human care, not a replacement.
Is AI therapy legal?
Increasingly regulated. In 2025 Illinois became the first US state to ban AI from delivering therapy while allowing administrative uses, Nevada and Utah passed restrictions, and six more states had bills advancing by early 2026. Using a chatbot for personal support remains legal everywhere; the laws target AI being provided as therapy.
Sources
- RAND Corporation: Nearly 1 in 5 US adolescents and young adults use AI chatbots for mental health advice. rand.org
- RAND / JAMA Pediatrics: AI chatbot use and disclosure for mental health among US adolescents. rand.org
- APA: 2026 Chatbots and Mental Health Survey. apa.org
- Psychology.com: AI therapy statistics roundup (KFF, OpenAI, Stanford, HRSA data). psychology.com
- Brilo: AI in mental health statistics (KFF, SAMHSA, Dartmouth, state legislation). brilo.ai
- CNN Health: Nearly 20% of young people use AI chatbots for mental health help. cnn.com
- HealthDay / US News: Youth AI mental health use coverage. usnews.com
- CognitiveFX: Mental health AI chatbot user survey. cognitivefxusa.com
- De Freitas et al.: AI Companions Reduce Loneliness, Journal of Consumer Research (via Prinsessa). prinsessa.com