Teens turn to AI when they're struggling, and those who use it for emotional support show worse mental health

By KRCG (The National News Desk) · October 28, 2026.

The Jed Foundation (JED), a US youth mental health organization, has published a report based on surveys of thousands of middle and high school students. According to its data, more than 80% use AI in some way. JED found no link between overall AI use and poorer mental health. It did find one for specific uses: schoolwork, advice on problems or relationships, companionship and emotional support.

The figures, according to the report itself: 58% use AI for schoolwork. More than a quarter use it to seek advice on relationships or problems. 12% use it for friendship or companionship. 57% turn to it when they feel sad, stressed, lonely or overwhelmed, but only 8% say they have sought specific emotional support from AI at those moments.

The groups with the strongest associations are as follows. Those who use it for homework report more loneliness and a greater history of suicidal thoughts. Those who seek advice on relationships or problems report more anxiety, depression, loneliness and suicide risk. Those who use it for companionship, or when they are struggling emotionally, also show a history of suicide attempts and lower resilience and well-being ("flourishing," in the report).

Kayla Anderson, JED's vice president of research and the report's author, stresses that this is an association, not causation. She considers it "very plausible" that young people who are already struggling lean more on AI. But she acknowledges that use could also worsen their problems. Determining AI's real role will, she says, be an important line of future research.

Our take: the easy headline would be "AI harms teenagers." The data do not say that. The most solid finding is a different one: AI already serves as a first resort in vulnerable moments, even, according to Anderson, when young people distrust it. That behavior exists, whatever its effect. The gap between the 57% who use it when feeling bad and the 8% who explicitly seek emotional support suggests use that is often informal, almost reflexive. That is hard to measure and to regulate.

The homework finding should also be read with caution. The fact that school use is associated with more loneliness does not mean that help with an essay isolates anyone. It may reflect an underlying variable, such as time spent alone or the student's profile. It is a reminder that a survey measuring only associations cannot separate cause from effect.

JED's recommendation is practical: don't let young people use AI as a friend or companion, talk early and often about what they use it for, and avoid offloading too much cognitive and relational work onto it. Anderson adds that the problem is also one of design: tools should be built with safety as a requirement, not as a patch.

Overall, in the short term this is the most sensitive ground in this technological transition. Conversational systems are always available and non-judgmental, and that draws in those who feel lonely. That same availability demands serious safeguards: crisis detection, referral to real people and transparency about their limits. Asking for evidence before legislating is sensible, but waiting for complete evidence is not, because the use is already happening.

In the long term, well-designed AI could expand access to psychological support where it does not reach today. But that horizon can only be reached if it is built on evidence, and this report is a good starting point precisely because it acknowledges what it does not yet know.