- In a nationally representative survey, 19.2% of adolescents and young adults ages 12 to 21 reported using AI chatbots for mental health advice in 2025.
- Nearly two‑thirds of participants (63.3%) indicated they had not shared their AI chatbot usage for mental‑health guidance with any other person.
- The authors recommended that mental‑health clinicians regularly ask patients about their use of AI chatbots.
A recent cross‑sectional, nationally representative survey found that roughly one in five young people had used an AI chatbot for mental‑health advice in the previous year.
Among individuals aged 12 to 21, 19.2% reported using AI chatbots for mental‑health advice in 2025, an increase from 13.1% in a comparable 2024 survey, according to Ryan K. McBain, PhD, MPH, of the RAND Corporation in Arlington, Virginia, and co‑authors in JAMA Pediatrics.
Among those who used AI chatbots for this purpose, 42.8% engaged with the tools at least monthly, and 91.7% rated the advice as somewhat or very helpful. However, 63.3% of participants indicated they had not discussed their AI chatbot usage for mental‑health advice with anyone.
“The rapid growth is striking, but even more noteworthy is that most young people who rely on these tools for mental‑health guidance say they keep that information to themselves,” McBain said in a press release.
The study “highlights the urgency of understanding and shaping the evolving role of AI chatbots in youth mental health care,” the authors wrote. “As these technologies become increasingly woven into the daily lives of young people, they should be recognized as active contributors to the broader ecosystem of psychological interventions.”
Co‑author Hao Yu, PhD, of Harvard Medical School in Boston, noted several perceived advantages of chatbots for mental health, including relentless availability, constant patience, and access in settings where mental‑health professionals are scarce. Nonetheless, “these tools can be hazardous,” Yu told MedPage Today.
The core issue is that chatbots are not designed as mental‑health instruments; they are not intended to diagnose or treat conditions and may disseminate both erroneous and misleading information, Yu explained.
Furthermore, the swiftly evolving AI environment possesses scant, if any, protective measures. Last year, the parents of a 16‑year‑old named Adam Raine filed a lawsuit against OpenAI and its CEO, alleging that ChatGPT contributed to the boy’s suicide.
Adolescents can pose any question to a chatbot, and if they display signs of suicidal ideation, the bot may inadvertently reinforce such behavior rather than discourage it, Yu warned.
Regarding the study’s implications, “I would anticipate that mental‑health clinicians will soon routinely ask patients, ‘Have you ever used AI chatbots for mental health concerns?'” Yu added.
The survey revealed that AI chatbot use for mental‑health advice was more prevalent among females than males (adjusted OR 2.10, 95% CI 1.36‑3.23), among those aged 18 to 21 compared with those aged 12 to 14 (aOR 3.65, 95% CI 1.98‑6.74), and among respondents who had consulted a physician about their mental health within the prior six months versus those who had not (aOR 1.89, 95% CI 1.18‑3.03).
Young adults’ higher engagement with chatbots may reflect a greater incidence of mental‑health conditions, broader smartphone access, and reduced parental oversight as children mature, McBain and colleagues noted.
Interestingly, youths who reported using AI chatbots for mental‑health support at least monthly were five times more likely to identify as Black (aOR 5.45, 95% CI 1.44‑20.66) than as White. The authors hypothesized that “Black youth may perceive professionals as less attuned to their unique needs,” or that heightened reliance on chatbots could indicate “reduced access to professional services.”
For this research, the authors recruited participants from RAND’s American Life Panel. The survey was administered to adolescents and young adults aged 12 to 21 in November 2025, encompassing a U.S. population‑weighted 42,825,655 youth (unweighted 1,009 youth). The median age was 17 years; 50% were male; 49.8% identified as White, 25.7% as Hispanic, and 13% as Black.
The study’s relatively modest sample size and a 58.4% response rate suggest that findings “contain uncertainty and may be subject to nonresponse bias,” the authors cautioned.
McBain noted that the survey did not assess the quality of care delivered by chatbots, which could be explored in future research. Yu highlighted another open question: why certain populations appear to rely on chatbots more heavily than others.
