Psychological pathways linking generative AI dependency to psychological capital depletion among Chinese international students

With the widespread use of generative AI in academic writing, language processing, and learning-support contexts, GAD has continued to rise. This study examines whether GAD is associated with rPsyCap and tests the mediating roles of PSD and rSE, both as independent mediators and as a sequential mediation mechanism. A three-wave longitudinal study was conducted. Using a combination of convenience and snowball sampling, questionnaire data were collected from 502 Chinese international students studying abroad (e.g., the United States, the United Kingdom, Australia, Canada, and South Korea). Participants completed measures of GAD, PSD, rSE, and rPsyCap. SEM was conducted using SPSS 26.0, AMOS 26.0, and CEFAtool 3.04. Direct and indirect effects were evaluated with 5,000 bootstrap resamples, and estimands for mediation and serial mediation were computed using AmosEstimandVB. GAD was positively associated with rPsyCap ( γ = 0.163, 95% CI [0.065, 0.264]). The sequential indirect effect via PSD → rSE was significant ( γ = 0.019, 95% CI [0.008, 0.036]), and PSD and rSE also showed significant independent mediation effects ( γ = 0.087, 95% CI [0.051, 0.132]; γ = 0.034, 95% CI [0.013, 0.060], respectively). GAD is linked to rPsyCap both directly and indirectly. By delineating a psychological process through which GAD may incur latent costs, this study informs the judicious integration of generative AI in educational and learning contexts.