Association between generative AI dependence and medical students’ clinical reasoning: the statistical mediating role of critical thinking and the moderating role of AI literacy

Generative artificial intelligence (GenAI) is rapidly transforming medical education. However, its educational consequences depend not only on the extent of use but also on whether students develop excessive dependence on AI-generated outputs. Based on cognitive offloading theory, this study develops a theoretical model examining the relationship between generative AI dependence and clinical reasoning, with critical thinking as a mediating mechanism and AI literacy as a boundary condition. A three-wave time-lagged survey design was adopted. Generative AI dependence and AI literacy were measured at Time 1, critical thinking was assessed at Time 2, and clinical reasoning was evaluated at Time 3. The results showed that generative AI dependence was negatively associated with both clinical reasoning ( β = −0.320, p < 0.001) and critical thinking ( β = −0.379, p < 0.001). Critical thinking positively predicted clinical reasoning ( β = 0.345, p < 0.001). After controlling for AI literacy, the indirect effect of generative AI dependence on clinical reasoning through critical thinking was −0.125, with a 95% confidence interval of [−0.165, −0.090]. Furthermore, AI literacy positively moderated the relationship between generative AI dependence and critical thinking ( β = 0.168, p < 0.001). The index of moderated mediation was 0.058, with a 95% confidence interval of [0.035, 0.085]. These findings suggest that medical education should focus not simply on restricting AI use but on promoting appropriate patterns of AI engagement. Developing AI literacy, output verification skills, and critical evaluation abilities may help students maintain effective human–AI collaboration while preserving independent clinical reasoning.