Artificial Intelligence-Assisted Feedback For Improving EFL Grammatical Accuracy

The expanding use of artificial intelligence (AI) in English as a foreign language (EFL) classrooms has made automated grammar correction increasingly accessible, but more accurate edited texts do not necessarily demonstrate lasting grammatical learning. This critical narrative review synthesizes 25 research and policy sources on written corrective feedback, automated writing evaluation, and generative AI to determine when AI-supported feedback can improve grammatical accuracy and when its benefits remain uncertain. The synthesis distinguishes error detection and immediate revision from learners’ subsequent independent performance. Research supports the potential of prompt feedback, repeated revision, and explanatory interaction; however, meta-analytic results show that gains in grammar may be smaller than gains in broader writing outcomes. Learner engagement, feedback reliability, grammatical feature, proficiency, and teacher mediation help explain divergent findings. A five-stage classroom framework is proposed: diagnose recurrent errors, elicit learner hypotheses, inspect AI suggestions critically, undertake controlled and meaningful revision, and assess transfer to new writing. The article concludes that AI is most defensible as a supervised formative-feedback resource, not an autonomous grammar teacher. Implications for assessment, equity, privacy, and future classroom research are discussed.