The serial mediating roles of AI self-efficacy and academic procrastination in the association between AI literacy and AI dependence among Chinese undergraduates

Objective Generative artificial intelligence is now embedded in university learning, but the psychological pathways linking AI literacy with AI dependence remain unclear. This study examined whether AI self-efficacy and academic procrastination mediate the association between AI literacy and AI dependence among Chinese undergraduates.Method A total of 417 Chinese undergraduate students aged 18–24 years completed measures of AI literacy, AI self-efficacy, academic procrastination and AI dependence. This study tested common method bias, estimated descriptive statistics and Pearson correlations, examined multicollinearity, conducted robustness checks and estimated a serial mediation model with PROCESS Model 6.Results AI literacy was negatively associated with AI dependence (r = –.283). Bootstrapped indirect effects indicated three significant pathways: through AI self-efficacy, through academic procrastination and through the sequential pathway from AI self-efficacy to academic procrastination. The total standardised indirect effect was −.229, accounting for 55.84% of the total effect.Conclusions AI self-efficacy and academic procrastination partially and sequentially mediated the association between AI literacy and AI dependence. The findings suggest that AI literacy may reduce maladaptive AI dependence by strengthening efficacy beliefs and reducing procrastination-related reliance on AI tools. Given the cross-sectional design, longitudinal research is needed to clarify temporal ordering.