From GenAI Use to L2 Writing Performance: A Moderated Sequential Mediation Model of Self-Efficacy and Metacognitive Strategy Use

Generative artificial intelligence is increasingly used to support second language writing; however, the motivational and strategic mechanisms linking GenAI use to L2 writing performance remain under-examined, particularly in exam-oriented contexts. Drawing on social cognitive theory and growth mindset theory, we tested a moderated sequential mediation model linking GenAI use to L2 writing performance through writing self-efficacy and metacognitive strategy use, with growth mindset moderating the relationship between GenAI use and self-efficacy. Survey data from 262 Chinese undergraduates who had completed the CET-4 were analyzed using partial least squares structural equation modeling (PLS-SEM) with 5,000 bootstrap resamples. The results showed that GenAI use was positively associated with self-efficacy, metacognitive strategy use, and L2 writing performance. Moreover, self-efficacy and metacognitive strategy use jointly mediated the relationship between GenAI use and writing performance, both independently and sequentially. Growth mindset further moderated the association between GenAI use and self-efficacy. These findings conceptualized GenAI as an environmental affordance that related to L2 writing outcomes through a motivational–strategic pathway, underscoring the importance of fostering growth-oriented beliefs and strategic regulation in GenAI-supported exam preparation.