Empowerment over enforcement: unpacking the psychological drivers of AI-assisted deep revision in EFL writing.

The rapid integration of generative artificial intelligence (AI) in higher education highlights a critical tension between top-down administrative enforcement and students' genuine cognitive investment. Drawing upon Self-Determination Theory (SDT) and Cognitive Load Theory (CLT), this study investigates how AI Prompting Literacy (APL) and External Mandate (EM) differentially predict the Deep Revision Engagement (DRE) of English as a Foreign Language (EFL) learners. Utilizing survey data from 327 Chinese university students and partial least squares structural equation modeling (PLS-SEM), we tested a multiple parallel mediation model incorporating Perceived Competence (PC), Intrinsic Motivation (IM), and Psychological Safety (PS). The empirical results demonstrate that External Mandate yields a negligible direct path, suggesting that compliance-driven requirements may fail to directly stimulate deep cognitive participation. Conversely, AI Prompting Literacy emerges as a significant predictor within this structural model. Higher prompting literacy exhibits a positive indirect association with learners' core psychological needs, thereby predicting an elevated level of Deep Revision Engagement through the parallel pathways of Perceived Competence, Intrinsic Motivation, and Psychological Safety. This research complements traditional technology acceptance frameworks by shifting the analytical lens toward internal psychological needs, suggesting that educational policies are more effective when focusing on cultivating student prompting capabilities alongside necessary institutional guidance.