Beyond grammar: How ChatGPT-4.0 shapes writing self-efficacy and performance in CSL writing

Abstract

This study examined ChatGPT-4.0 as a feedback tool in Chinese as a Second Language writing instruction, focusing on writing performance and writing self-efficacy. An eight-week quasi-experimental intervention was conducted with 53 intermediate-to-advanced CSL undergraduates in East China. The experimental group received ChatGPT-assisted feedback during writing, while the control group relied on traditional teacher support. A mixed-methods design was adopted, combining pre- and post-test writing assessments, questionnaire measures of self-efficacy, and follow-up interviews. Findings indicated that students using ChatGPT showed significantly higher post-test writing scores and greater gains in writing self-efficacy, particularly in ideation, conventions, and self-regulation. Regression analyses further suggested that ChatGPT use and post-test self-efficacy were both associated with improved writing outcomes. Interview data revealed that learners valued the immediacy, availability, and low-pressure nature of AI feedback, which appeared to support correction, idea development, self-monitoring, and confidence building, while also raising concerns about overreliance and formulaic expression. Overall, the results suggest that ChatGPT may serve as a useful feedback resource in CSL writing when implemented with structured guidance, teacher mediation, and attention to learner uptake of AI-generated feedback.