Enhancing L2 writing with AI: Effects of ChatGPT feedback and the role of student engagement

Abstract

The rapid development of Generative AI (GAI) tools offers new opportunities for providing immediate and personalized feedback in writing instruction. While previous studies have explored its potential as an automated written corrective feedback (AWCF) tool, little is known about how effective its feedback is across different aspects of L2 writing, and how such engagement influences writing outcomes. The present study examined the effects of ChatGPT feedback on undergraduate students’ L2 writing performance, including lexis, grammar, content and organization, and investigated the role of learner engagement, including behavioral, cognitive, emotional engagement, in influencing these effects. A total of 273 students with comparable English proficiency were divided into two groups: one received teacher feedback while the other obtained ChatGPT feedback. Results showed that ChatGPT feedback led to significant improvements in grammar and lexis, but had limited impact on content and organization. In contrast, the teacher group did not show significant improvement in any writing dimensions. Analysis of student engagement revealed that all three dimensions were positively associated with writing improvement. Among them, behavioral engagement was the strongest and most direct predictor of the effectiveness of ChatGPT feedback. Practical implications for using ChatGPT in writing instruction and directions for future research are discussed.