Use of ai-based mental health tools and psychological well-being among Chinese university students: a parallel mediation model of emotional self-efficacy and perceived autonomy.

While AI-powered mental health tools have gained momentum as accessible and user-centered alternatives to traditional services, the psychological processes through which they exert influence remain insufficiently theorized. This gap is especially relevant in contexts such as China, where cultural norms surrounding mental health stigma and reluctance toward emotional disclosure intersect with shifting generational attitudes toward autonomy and self-management. This study aimed to examine the relationship between AI-based mental health tools and psychological well-being among Chinese university students, specifically focusing on the mediating roles of emotional self-efficacy and perceived autonomy in mental health management within a parallel mediation framework. A cross-sectional survey was administered to 3,859 university students across Jiangxi Province, China. Participants completed validated measures assessing use of AI-based mental health tools, emotional self-efficacy, perceived autonomy, and psychological well-being. Structural equation modeling was employed to test a parallel mediation model, controlling for age, gender, academic level, and prior use of mental health apps. A post hoc sensitivity analysis was also conducted to examine the robustness of mediation effects to potential unmeasured confounding. The results showed that using AI-based mental health tools was positively associated with psychological well-being (β = 0.229, p < 0.001). Both emotional self-efficacy (β = 0.121, 95% CI: 0.101,0.159) and perceived autonomy in mental health management (β = 0.132, 95% CI: 0.112,0.176) significantly and partially mediated this association, indicating that the psychological benefits of tool use were transmitted through enhanced emotional regulation and volitional engagement. Bootstrapped indirect effects were statistically significant, with confidence intervals that did not include zero. The overall model explained 38% of the variance in psychological well-being. Sensitivity analysis further demonstrated that the mediation effects were robust to moderate levels of unmeasured confounding. The findings suggest that AI-based mental health tools may enhance student well-being by facilitating symptom management and strengthening emotional self-regulation and volitional engagement in care. These results underscore the importance of autonomy-supportive and empowerment-focused design in digital mental health interventions. Future longitudinal and cross-cultural research is recommended to validate these pathways and inform scalable, context-sensitive applications.