Reading–Writing Integration Through ChatGPT-Supported Feedback: Growth, Associations, and Mediated Pathways of Student Engagement in Chinese Classrooms

Abstract Although reading–writing integration has been increasingly advocated in Chinese language education, limited empirical evidence explains how reading, writing, and student engagement become integrated, particularly under AI-supported feedback. To address this issue, this study examined whether ChatGPT-supported feedback improved reading outcomes, writing outcomes, and student engagement, while investigating their interrelationships, the mediating role of student engagement, and the corresponding pathways in reading–writing integration. A total of 400 middle school students from eight intact classes in Shenzhen participated, with 200 students receiving ChatGPT-supported feedback and 200 receiving human-generated feedback. Using a quasi-experimental pretest–posttest design, data were analyzed within a quantitative longitudinal framework integrating baseline equivalence tests, 2 × 2 mixed-design ANOVAs, factor-level correlation analyses, and multi-group cross-lagged path analysis to examine changes and temporal relationships among reading, student engagement, and writing across two groups (N = 400). Despite comparable pretest scores, findings showed that Group 1 demonstrated significantly greater pretest–posttest gains in reading, student engagement, and writing, with significant Time × Group interactions and large effects. Correlations further indicated stronger posttest associations among engagement, reading, and writing in Group 1, whereas these relationships remained weak in Group 2. Longitudinal analyses revealed stronger T2 within-time direct and mediating pathways in Group 1, alongside several significant cross-lagged effects, but cross-lagged mediating effects were generally nonsignificant in both groups. The findings theoretically highlight student engagement as a potentially time-sensitive mechanism through which AI-supported learning may strengthen the integration of reading and writing. Practically, they suggest the pedagogical value of AI-supported learning environments for fostering engagement and connecting literacy processes, while emphasizing sustained instructional exposure to consolidate these relationships over time.