Generative AI Dependency and University Students’ Mental Health: Examining the Mediating Role of Academic Stress and Psychological Well-Being

Abstract

Generative artificial intelligence (GenAI) is now a routine part of university learning, but the psychological consequences of becoming dependent on these tools are not yet fully understood. This study tested a parallel-mediation model linking GenAI dependency with university students’ mental-health difficulties through academic stress and psychological well-being. The framework draws on the Interaction of Person–Affect–Cognition–Execution model, cognitive-offloading theory, stress-and-coping theory, and self-determination perspectives, while distinguishing purposeful AI use from dependency marked by cognitive reliance, reduced control, and lower confidence in completing academic tasks independently. We analyzed data from 420 university students in IBM SPSS Statistics using reliability analysis, descriptive statistics, Pearson correlations, regression analysis, and Hayes’ PROCESS Model 4 with 5,000 bootstrap samples. GenAI dependency was positively associated with academic stress and mental-health difficulties and negatively associated with psychological well-being. Academic stress and psychological well-being both carried significant indirect effects between GenAI dependency and mental-health difficulties, while the direct effect remained significant, indicating partial mediation. The findings suggest that psychological risk is linked less to AI use itself than to patterns of maladaptive reliance. Universities can respond by strengthening AI literacy, self-regulated learning, authentic assessment, manageable academic demands, and access to mental-health support. Because the study is cross-sectional, interpret the results as associations rather than evidence of causal direction.