A study of nearly 40,000 Ontario schoolchildren found that those who turned to AI chatbots for emotional support were roughly twice as likely to meet criteria for clinical-level emotional problems as those who did not, according to researchers at the University of Ottawa. The study, titled “Affective Generative Artificial Intelligence Use and Youth Mental Health,” was published Aug. 31 in JAMA Pediatrics. Among students who sought emotional support from AI, 57.7% met the study’s threshold for serious emotional difficulty, compared with 29.2% of students who did not use AI that way.
The research team drew on survey responses collected between May 2025 and March 2026 across four Ontario school boards, part of the ongoing Ontario Health and Peer Relations Study. Researchers say the pattern is best read as a warning sign worth noticing, not evidence that chatbots themselves are causing the distress they found.
Nearly 40,000 Ontario students surveyed
Researchers examined responses from 39,761 students in grades 4 through 12. Of those, 8,408 students — about 21% — said they had used an AI chatbot for help managing their feelings or for advice about relationships with friends, family or others. That tendency was more common among older students and among racialized and gender-diverse youth compared with their white or male peers, the university reported.
The surveys, completed anonymously in schools, covered AI use alongside measures of emotional health, loneliness and a young person’s sense of mattering to others. Researchers used statistical models built to account for students being clustered within particular schools and for missing survey answers.
Bonnesen led that statistical analysis and wrote up the results, working alongside Vaillancourt, who co-conceptualized the project. The two drew their sample from the Ontario Health and Peer Relations Study, an ongoing survey program spanning four school boards rather than a poll built specifically to study AI use, which is part of why the dataset already contained years of loneliness and mattering measures to compare against.
57.7 percent versus 29.2 percent: the size of the gap
The gap between the two groups held up even after researchers adjusted for demographics, loneliness, a student’s sense of mattering, and how much a student used AI for schoolwork. After those adjustments, using AI for emotional support was still linked to a 26% higher prevalence of clinical-level emotional difficulties. Students who leaned on AI this way also reported feeling lonelier and less valued by the people around them.
Because the study is cross-sectional, meaning it captured a single point in time rather than tracking students over years, it cannot establish whether distress drove students to AI or whether the AI use itself deepened that distress. Researchers were explicit that the design allows them to describe an association, not a cause.
Students who relied heavily on AI for schoolwork were also more likely to turn to it for emotional support, the university’s release noted, and the two behaviors overlapped often enough that researchers had to check whether one was simply standing in for the other. The 26% adjusted gap held even after accounting for that overlap, meaning heavy homework use of AI does not by itself explain why the emotional-support group showed more distress.
Kamilla Bonnesen on what the pattern does not prove
Kamilla Bonnesen, a postdoctoral fellow at the Brain and Behaviour Lab at the University of Ottawa and the study’s lead author, described the scale of the difference in plain terms. “Of almost 40,000 students, those who leaned on AI for emotional support were much more likely to be struggling with clinical-level emotional issues, more loneliness, and a lower sense of mattering,” Bonnesen said.
Bonnesen was careful to separate the finding from any claim about what causes what. “This doesn’t mean that AI is causing these struggles,” Bonnesen said. “Instead, using AI in this way might be a sign that a young person needs more human connection and support.”
Tracy Vaillancourt’s call to ask why, not just whether
Tracy Vaillancourt, Canada Research Chair in Youth Mental Health and Violence Prevention at the University of Ottawa’s Faculty of Education, co-conceptualized the study with Bonnesen and framed its practical use for the adults around a struggling child. “As chatbots become part of young people’s daily lives, parents, educators and mental health professionals shouldn’t just ask if youth are using AI, but why,” Vaillancourt said.
Vaillancourt added that the finding is best treated as an opening rather than a verdict. “When a student is turning to AI for emotional reasons, it’s a good opportunity to start an open, nonjudgmental conversation,” Vaillancourt said. “It is a chance to reconnect a struggling student with meaningful human support.” The University of Ottawa’s release credits Amanda Krygsman, Sarah Hobson and Daphne Korczak as additional co-authors alongside Bonnesen and Vaillancourt.
Krygsman, Hobson and Korczak rounded out the five-person author team anchored at the University of Ottawa’s Faculty of Education, contributing to the survey design and the analysis of loneliness and mattering scores, according to the university. Whether the same pattern holds as chatbots grow more capable, or as today’s fourth-graders reach their teenage years, is a question the research team has not yet answered.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- A handful of car transmissions are so tough that mechanics say they almost never die
- Card skimmers hidden on gas pumps are draining accounts, and there’s a quick way to spot them
- The NSA warns one messaging setting can clone your texts to a stranger
- Security experts still urge phone owners to switch off one location-tracking setting
