Seeking AI support may flag emotional issues

Youth who turned to generative AI for feelings or interpersonal advice were more likely to meet a clinical symptom threshold.

By
Andrea Surnit
October 12, 2026
4 min

MDSpire News

Overview

In a cross-sectional analysis of 39,761 Ontario students in grades 4 through 12, use of generative AI for emotional support or interpersonal advice was associated with a higher prevalence of clinically significant emotional issues. The association persisted after adjustment for measured demographic and psychosocial factors, but the study cannot establish whether emotional distress preceded or followed affective AI use.

Background

Generative AI is used by young people for both schoolwork and personal concerns, but these uses may have different relationships with emotional health. In this study, affective use meant asking an AI chatbot for help with feelings or relationship advice, while functional use referred to schoolwork. Emotional issues were assessed with seven items from the Ontario Child Health Study Emotional Behavioral Scales and a validated cutoff for clinically significant symptoms. The supplied context also describes pediatric professional guidance emphasizing child-specific design, transparency, validation, privacy, and safeguards for AI tools used with minors.

Data Highlights

Measure	Finding
Study population	39,761 students in grades 4–12; mean age 12.8 years; four Ontario school boards
AI use	21% reported affective AI use; 43% reported functional AI use at least once
Clinically significant emotional issues	58% among affective AI users vs. 29% among nonusers
Adjusted association	Affective AI use remained associated with a 26% higher prevalence after adjustment for demographic characteristics, loneliness, mattering, and functional AI use
Functional AI use	Emotional-problem score increases ranged from 0.01 for a little use to 0.04 for a lot of use, compared with never using AI for schoolwork
Affective AI use by group	27% of female, 14% of male, and 34% of gender-diverse students; prevalence increased from 16% in elementary to 21% in middle and 26% in secondary school

Key Findings

Clinically significant emotional issues were reported by 58% of students who used AI for emotional support or interpersonal advice, compared with 29% of nonusers.
After adjustment for demographic characteristics, loneliness, perceived mattering, and functional AI use, affective AI use remained associated with a 26% higher prevalence of clinically significant emotional issues and with higher continuous emotional-problem scores.
Functional AI use showed a smaller dose-response association: emotional-problem scores increased by 0.01 among students reporting a little school-related AI use and by 0.04 among those reporting a lot, relative to students reporting no use.
Affective AI use was reported by 27% of female students, 14% of male students, and 34% of gender-diverse students; it was more common at higher school levels and among Black, Indigenous, and other racial and ethnic groups than among White students (24% vs. 18%).
The association between affective AI use and emotional issues was attenuated among gender-diverse students compared with female and male students; investigators found no difference between female and male students. School level did not modify the association.
The cross-sectional design, self-reported measures, and measurement of affective AI use as use versus nonuse limit conclusions about temporal sequence, causality, and frequency or duration of use.

Clinical Implications

The study authors wrote that clinicians should normalize screening for affective AI use because they considered it a robust marker for underlying mental health concerns. The findings do not establish that AI use causes emotional issues, and the authors noted that the results concern general-purpose AI and may not generalize to specialized clinical AI agents.

Conclusion

In this Ontario student sample, affective AI use was more strongly associated with emotional issues than functional AI use, including after adjustment for measured factors. Longitudinal research would be needed to clarify the direction and course of this association.