The Effectiveness of AI-based Teacher Feedback in Enhancing Textual Coherence in EFL Writing

Abstract

Academic writing remains a challenge for Thai university students, particularly in achieving textual coherence beyond surface-level grammar. This study investigates the effectiveness of AI-assisted teacher feedback—leveraging generative models like ChatGPT and Gemini—in enhancing the use of cohesive devices and thematic progression patterns among Thai graduate students. Utilizing a quasi-experimental, one-group pre-test/post-test design, the research involved 23 business graduate students (A2–B1 proficiency) enrolled in a remedial English course. Participants produced expository cause-effect paragraphs and received directive, AI-informed teacher feedback through Google Classroom. Data were analyzed using paired-sample t-tests and semi-structured interviews to triangulate writing performance with student perceptions. Quantitative results revealed a statistically significant improvement in overall textual coherence (t = 26.72, p < 0.05), with specific gains in the mastery of Theme-Rheme structures and a more diverse repertoire of cohesive ties. Qualitative findings indicated that students perceived the feedback as a vital cognitive guidance that fostered metalinguistic awareness and facilitated a smoother information flow. The study concludes that while AI provides consistent, systematic data for feedback, its integration with Systemic Functional Linguistics (SFL) frameworks allows students to transition from incoherent prose to unified academic discourse. These findings imply that a hybrid pedagogical approach—combining AI efficiency with explicit thematic instruction—can significantly promote textual coherence in L2 writing. This study contributes original value by bridging the gap between automated feedback technology and macro-structural discourse theories in an EAP context.