This cross-sectional survey examined whether AI feedback literacy and revision behaviour mediated the association between large language model (LLM) co-editing intensity and writing self-efficacy among 305 Vietnamese undergraduates with recent experience using LLM tools for English writing revision. The questionnaire measured LLM co-editing intensity, AI feedback literacy, revision behaviour, and writing self-efficacy. Reliability estimates were acceptable to good, and exploratory factor analysis supported the proposed four-factor structure. Serial mediation analysis with 5,000 bootstrap resamples showed significant indirect associations between LLM co-editing intensity and writing self-efficacy through AI feedback literacy, revision behaviour, and their sequential pathway. The serial indirect effect was 0.048, with a 95% bootstrap confidence interval of [0.024, 0.080], and the total indirect effect was 0.206, 95% CI [0.140, 0.283]. Given the cross-sectional and self-reported data, these findings are interpreted as associational rather than causal. Pedagogically, the results suggest that LLM-supported EFL writing instruction should focus not simply on increasing tool use but on helping students evaluate AI-generated feedback and translate it into purposeful revision decisions.
Copyrights © 2026