A JAMA Pediatrics study finds adolescent AI chatbot use for mental health advice jumped more than 40% in a year, outpacing the clinical, parental, and regulatory frameworks meant to protect minors.
Roughly one in five U.S. adolescents and young adults now turn to AI chatbots for mental health advice, up from about one in eight a year earlier. The pace of adoption has outrun the safety frameworks meant to govern it, and that gap, rather than the technology itself, is what regulators, clinicians, and parents are now trying to close.
The numbers come from a study published in JAMA Pediatrics by researchers at the RAND Corporation, drawing on survey data from U.S. adolescents and young adults. Per a RAND press release summarizing the findings, the recent wave shows about 1 in 5 respondents reporting mental-health chatbot use, compared with roughly 1 in 8 the prior year. That is a relative jump of more than 40% in twelve months, putting a behavior most adults still consider novel inside the routine of a generation.
The accompanying RAND commentary by the study's authors frames the moment bluntly: for millions of adolescents, AI is already part of the mental health landscape. The authors argue for guardrails calibrated to how minors actually use these tools: age-appropriate disclosure, crisis-routing that reaches humans, parental awareness without surveillance, and clinical accountability when a chatbot is positioned as a therapeutic aide. It is an argument, not yet a consensus, and the empirical paper underneath it does not measure whether those conversations actually help or harm.
The institutions meant to govern that landscape have not caught up. No binding federal rule requires mental-health chatbots to meet a clinical standard of care, to escalate crisis signals to a human, or to disclose that the user is talking to a synthetic interlocutor. No standard requires parental visibility into these conversations. The major consumer AI products operate under voluntary terms while federal regulators and Congress work out what, if anything, to mandate.
The first concrete moves are starting to appear. The Federal Trade Commission opened a formal inquiry in September 2025 into AI chatbots "acting as companions", examining whether existing consumer-protection law reaches platforms that simulate emotional relationships with minors. On Capitol Hill, a Senate proposal to ban AI companion chatbots for minors has advanced past an early procedural step, and the Youth AI Privacy Act (S.4199) sits on the public docket at Congress.gov. None of those has produced binding rules, but together they sketch the first U.S. attempt to define what a chatbot owes a 14-year-old in distress.
The limits of the data deserve attention. The JAMA Pediatrics study measures self-reported use; it does not measure clinical outcomes, harms averted, or harms caused. Coverage of the findings, including NBC Palm Springs' write-up, has tended to treat the 1-in-5 figure as a warning, which is fair to the adoption curve, but says little about whether the conversations themselves are net helpful or harmful. The growth is real. The question of whether the average teen chatbot conversation is making things better or worse is genuinely open, and that question will shape whether the rules land as labeling, as escalation duties, or as outright limits.
What to watch next: whether the FTC inquiry converts into concrete enforcement against specific products; whether the Senate companion-ban proposal reaches a floor vote in this Congress; and whether the JAMA Pediatrics researchers, or other groups, publish follow-up data on outcomes rather than only on usage. The adoption curve has already landed inside the mental-health system. The safety curve is still being written.