AI-assisted writing feedback in EFL: Tracking student performance and reflections

Abstract

This study explores how AI-assisted writing feedback supports English as a Foreign Language (EFL) learners' writing development and feedback literacy in a Japanese university. Twenty-one first-year students completed nine Write & Improve (W&I) tasks, progressing from descriptive to argumentative essays. Each task was revised following W&I feedback, and students were asked to write a short reflective log. An explanatory mixed-methods approach was adopted, combining quantitative analyses of writing performance and linguistic features with qualitative coding of students' reflections. Findings showed measurable gains in both higher- and lower-level groups, with particularly notable improvement among lower-level students in complexity, fluency, and sophistication. While the higher group consistently outperformed the lower group in Term 1, this gap narrowed in Term 2. Reflection logs provided important insights into feedback literacy: students initially valued W&I for surface-level corrections but later expressed frustration as tasks became more complex and scores plateaued. This tendency was especially evident among lower-level students, reflecting both the affordances and limitations of automated feedback. The study concludes that AI-assisted tools can foster writing development and emerging feedback literacy, but sustained progress requires teacher mediation. Integrating Automated Writing Evaluation (AWE) into ecological feedback environments - combining AI, teacher, and student reflection - offers a promising approach for sustainable L2 writing instruction.