Teens and young adults are turning to generative artificial intelligence (GenAI) for mental health support. Many of the conversations about AI center on its risks, which are real. But because youth are already using these tools, we should also ask: When is AI an endpoint, and when could it be a bridge to care?
GenAI has been documented giving instructions on lethal means of suicide, providing false or dangerous guidance, inventing fake therapy degrees, advising youth on how to hide symptoms from adults, and engaging in intimate conversations with minors. If these systems connected youth with real human support instead, they could help more young people find mental health care, instead of replacing it. .
To better understand how AI enters youth mental health pathways, Surgo Health and The Jed Foundation (JED) surveyed a nationally representative sample of over 1,300 youth ages 13-24, asking them about their use of AI tools for mental health support.
Twelve percent of the young people reported using GenAI for their mental health and most of those (69%) used tools not designed for that purpose.
Youth turn to AI not because it works well, but because other options feel unattainable .
Youth whose caregivers didn’t support them when they asked for help or who didn’t know it was possible to get professional help were between 2 to 3 times more likely to have used GenAI for mental health support.
Sixty-five percent reported turning to GenAI because they had no one they trusted to confide in. Those using GenAI were also more likely to report financial barriers and difficulties finding transportation. These barriers fell hardest on younger youth, boys, and Black and Hispanic youth. GenAI may become a fallback because it’s low-friction: anonymous, always available, free, and easier than sharing with people.
Despite its increasing adoption, youth are not finding GenAI helpful for mental health.
Many young people who used AI for mental health indicated that they knew AI was telling them what they wanted to hear. One survey participant said, “It’s free and easy to use; I don’t have to pay a penny for something to give me comforting words, even if they don’t really exist.” After talking to GenAI, the participant said he felt “A little better, but it felt like putting a band-aid on a giant gash.”
This experience is not unique. Half of the young people surveyed felt neutral or negative about the support they got from GenAI. One in three reported wishing they had not used it for this reason. The immediate appeal of short-term relief is too often quickly replaced by regret.
Approximately one-third of the youth who used GenAI relied on it as their only source of support. This group was more likely to report neutral or negative experiences than the group who used it before, after, or alongside professional support.
This pattern suggests GenAI’s limitations become more apparent, and potentially more harmful, when it’s the only support a young person receives.
A path forward — integrating offline and online worlds
Concerningly, in the research from Surgo Health and JED, approximately one in three youth who used GenAI for mental health concerns reported it did not prompt them to seek professional care. Ideally, AI would act as a bridge, pointing young people to the right care at the right time. To achieve this, we have to expand how these models recognize and intervene in moments of distress. AI systems are incentivized tokeep users engaged instead of connecting them to humans who can help, yet we need to ensure that online and offline worlds are set up to effectively relay the baton from one to the other. We also need to ensure that there are robust school-based services, affordable community care, and parents, caregivers, and other caring adults equipped to respond when youth are struggling.
While AI didn’t create the gaps in our mental health system, if left unchecked, it may deepen them. Online and offline worlds must be set up to work together, or the bridge leads to nowhere.