A Comparative Analysis of AI-Generated Feedback Tools on EFL Students’ Writing in Higher Education

Abstract

This research compared the feedback of three generative artificial intelligence (AI) tools, ChatGPT, Google Gemini, and Microsoft Copilot, on EFL students’ writing. It adopted a mixed-methods design to evaluate the writing of 35 medical students who enrolled in an Intensive English for Academic Purposes (IEAP) course at Taif University during 2025-2026. The AI-generated feedback was coded into five categories: grammar, vocabulary, cohesion, sentence restructuring, and mechanics. Quantitative data were analyzed through descriptive statistics and one-way repeated measures ANOVA, while qualitative data were examined through thematic analysis of feedback style and depth. The findings revealed that grammar was observed to be the most common feedback type in all three tools. Through a one-way repeated measures ANOVA test, there were no significant differences in the feedback on grammar and cohesion in all three tools. There were significant differences in the feedback on vocabulary, sentence restructuring, and mechanics. MS Copilot had a significantly higher number of vocabulary and sentence restructuring feedback than ChatGPT and Gemini, while ChatGPT had a significantly higher number of mechanics feedback. The qualitative analysis showed different types of feedback where ChatGPT worked as a corrective feedback tool, Gemini worked on instructional methods combining corrections with explanations and suggestions, and MS Copilot used text improvements through sentence reformulation. The research will contribute to the existing literature regarding AI-assisted writing by demonstrating that AI tools differ in the feedback they provide and in their pedagogical focus, style, and depth.