Generative AI as an IELTS Tutor: The effects of AI-generated feedback on EFL students’ writing

Abstract

This study examines the impact of AI-generated feedback on Thai university students’ performance in IELTS writing. Additionally, it explores the students’ perceptions of the feedback and their evaluations of its pedagogical value. Adopting a sequential mixed-methods research design, the study integrates quantitative and qualitative evidence collected from 52 participants enrolled in an English Preparation for IELTS at a public university in Thailand. The effectiveness of AI-generated feedback was evaluated through a comparative analysis of students’ pre- and post-intervention writing performance, while students’ perceptions of the feedback were explored using questionnaire data and semi-structured interviews. The findings indicate that engagement with AI-generated feedback led to improvements in students’ post-test writing performance across both writing tasks. The gains were statistically significant in three assessment criteria, with the most notable gains observed in grammatical range and accuracy. Despite marginal improvements, however, the gains on content-related dimensions were limited. The students perceived AI-generated feedback as beneficial and useful for their writing development, while still feeling the need for human feedback. The study concludes that while AI-generated feedback can effectively support the development of specific writing features, meaningful instructional impact continues to depend on active teacher involvement in the feedback cycle.