Most teens who used AI for mental health support wanted more help

By Vijay Kumar MalesuReviewed by Susha Cheriyedath, M.Sc.Oct 28 2026

A survey of more than 24,000 students in England examines who turns to AI chatbots for mental health advice and how their experiences fit into the wider search for support.

Study: General-purpose AI as a source of mental health support amongst 24,000 adolescents in England. Image Credit: Fabio Principe / Shutterstock

Among more than 24,000 students surveyed in England, 2.7% reported using general-purpose artificial intelligence (AI) for mental health support. More than half of those who rated their experiences found it helpful, but nearly two-thirds of users wanted additional assistance. The study, published in the journal npj Mental Health Research, examined who used these platforms and what other support they accessed or wanted.

Background

General-purpose AI and large language model (LLM) chatbots are increasingly used for mental health advice, raising questions about their benefits, limitations, and safety. Adolescence is a critical developmental stage during which mental health issues may first appear and intensify. Digital tools could offer an alternative or complement to conventional support, particularly when personal, cultural, or service-related obstacles limit access. Questions remain about clinical oversight, safety measures, and whether these tools are suitable. Further research is needed to systematically evaluate their benefits and potential harms.

About the study

The researchers analyzed data from the 2025 OxWell Student Survey, a repeated cross-sectional survey examining mental health and well-being among students attending secondary schools and further education colleges in England. School recruitment was nonprobabilistic and involved schools invited through local authorities and education groups. The survey was administered during school hours between February and May 2025, mainly in Oxfordshire and Liverpool. The nonrandom selection of schools and concentration in these areas mean the findings may not represent adolescents across England.

After exclusions and consent screening, 24,132 students reached the survey section assessing AI and LLM use for mental health support and were included in the analysis. The mean age was 14.0 years among participants aged 18 or younger.

The researchers assessed platform use in the past year, perceived helpfulness, unmet support needs, and whether participants accessed other forms of support. Predictors covered demographic characteristics, mental health, self-harm, and social adversity. Depression and anxiety symptoms were measured with a standardized questionnaire. The researchers used descriptive analyses and logistic regression to examine differences across participant groups.

Study results

Of 24,132 adolescents, 662 (2.7%) reported using general-purpose AI or LLM platforms for mental health support in the past year. Of respondents who rated their experiences, 55.9% found the support helpful, 31.4% found it neither helpful nor unhelpful, and 12.8% found it unhelpful. Among AI users who answered a separate question, 63.2% said they would have liked additional mental health support during the previous year. This was a measure of perceived unmet need and did not establish whether the AI support itself was inadequate.

Most adolescents who used these platforms had also accessed other forms of mental health support. Overall, 92.1% reported receiving support elsewhere, while 46.1% had accessed formal services during the previous year. Only 52 adolescents reported using AI and LLM platforms without other support. Another 57 used AI alongside other online resources but reported no offline support. Together, these groups accounted for 16.5% of AI users, who may be less visible to conventional services.

Several characteristics were associated with platform use. Transgender and gender-diverse adolescents had higher odds of use than the reference group of boys and young men. AI use was also more common among those reporting frequent food insecurity, higher depression and anxiety symptoms, previous self-harm, frequent loneliness, or occasional bullying. Adolescents from Asian or Asian British and other minoritized ethnic groups had higher odds of use. The strongest associations were observed in the Asian or Asian British and Other ethnic categories.

Perceived helpfulness showed few differences across participant groups. Asian or Asian British adolescents were the only group significantly more likely to rate the support helpful.

Perceived unmet support needs were more common among Asian or Asian British adolescents, those with higher depression and anxiety scores, and those reporting previous self-harm. Frequent loneliness was also associated with unmet needs. Adolescents reporting occasional bullying or intermittent food insecurity were also more likely to want additional support, but this pattern was not observed at higher levels of these adversities.

Conclusions

General-purpose AI and LLMs were used for mental health support by a small proportion of adolescents in this English survey. While more than half of respondents who rated their experiences considered the support helpful, most users had also accessed other support, and many wanted additional assistance. Platform use and unmet needs were concentrated among adolescents experiencing particular vulnerabilities, including mental health difficulties, self-harm, loneliness, and social adversity.

These findings suggest that AI is often used alongside existing support systems but cannot be assumed to meet adolescents' needs adequately. The study did not assess whether AI advice was clinically appropriate, safe, or effective in improving mental health. The 2.7% estimate may be lower than estimates from other studies because of how mental health support was defined, while actual use may have changed since the 2025 survey.