Modelling AI-mediated writing feedback in EFL: the role of ChatGPT use, feedback quality, motivation, perceived writing improvement, and perceived assessment performance

This study examines how ChatGPT-generated writing feedback relates to Saudi EFL learners’ perceived assessment performance among those who already use ChatGPT, and how this relationship is associated with indirect statistical pathways involving evaluative and motivational factors. Drawing on the Feedback Engagement Framework and Self-Regulated Learning Theory, a structural model was specified to account for relationships among ChatGPT use, perceived feedback quality, writing confidence and motivation, perceived improvement in writing skills, and perceived assessment performance. Data were collected from 662 undergraduate learners who reported prior use of ChatGPT and analysed using structural equation modelling. The results indicate that ChatGPT use does not directly predict perceived assessment performance. Instead, its association with perceived assessment performance was indirect, through statistical pathways involving perceived feedback quality, confidence and motivation, and perceived improvement in writing skills. Perceived feedback quality was strongly associated with both motivational responses and perceived improvement in writing skills. Academic level moderated the relationship between feedback quality and perceived improvement in writing skills, whereas gender did not. The findings indicate that learners’ evaluations of AI-generated feedback are associated with their perceptions of improvements in writing skills and assessment performance, rather than with objectively measured learning outcomes.