Generative AI dependency and self-perceived clinical decision-making in nursing students: A three-wave longitudinal study of metacognitive awareness as a mediator.

BACKGROUND: Generative artificial intelligence tools are increasingly used in nursing education, but excessive reliance on externally generated reasoning may be associated with reduced cognitive engagement. Evidence remains limited on whether within-person changes in generative artificial intelligence dependency are prospectively associated with nursing students' metacognitive awareness and self-perceived clinical decision-making. OBJECTIVES: To examine longitudinal within-person associations among generative artificial intelligence dependency, metacognitive awareness, and self-perceived clinical decision-making, and to test whether metacognitive awareness mediates these associations. DESIGN: A three-wave prospective longitudinal panel study. SETTINGS: Nursing schools at three medical universities in China. PARTICIPANTS: A convenience sample of 687 undergraduate nursing students completed the baseline survey; 618 students completed the third wave, yielding a retention rate of 90.0%. METHODS: Data were collected at the beginning, middle, and end of one academic semester. A random-intercept cross-lagged panel model was used to separate stable between-person differences from within-person fluctuations. Missing data were handled using full-information maximum likelihood. Common method bias was examined using Harman's single-factor test and an unmeasured latent method construct. RESULTS: At the within-person level, higher generative artificial intelligence dependency was prospectively associated with lower metacognitive awareness at the subsequent wave (β = -0.18 to -0.21, p < .001). Higher metacognitive awareness was prospectively associated with higher self-perceived clinical decision-making (β = 0.22 to 0.25, p < .001). The indirect pathway from dependency to self-perceived clinical decision-making through metacognitive awareness was significant (standardized indirect effect = -0.045, 95% bootstrap confidence interval [-0.074, -0.019]), whereas the direct pathway was not statistically significant. CONCLUSIONS: Generative artificial intelligence dependency was prospectively associated with students' self-perceived clinical decision-making through metacognitive awareness. Because the study was observational and the outcome was self-reported, the findings indicate an educational risk pathway rather than demonstrated clinical harm. Nursing educators may benefit from pairing artificial intelligence use with metacognitive scaffolding, verification routines, and reflective assessment.