Chatbot as a Feedback Partner: An Applied Linguistics Study on AI-Mediated Corrective Feedback in Enhancing EFL Learners' Writing Accuracy and Revision Behavior

This study investigates the role of AI chatbot-generated corrective feedback in enhancing the writing accuracy and revision behavior of EFL undergraduate learners. Employing a descriptive qualitative design grounded in Applied Linguistics, the research involved undergraduate students who exhibited persistent grammatical errors, limited revision strategies, and infrequent access to individualized writing feedback. Data were gathered through multiple drafts of argumentative paragraphs, chatbot-generated feedback logs, and semi-structured interviews conducted across several feedback-revision cycles. The findings indicate substantial improvement across four indicators: grammatical accuracy, depth of revision, error noticing, and learner autonomy in the editing process. Learners progressively shifted from surface-level corrections, such as spelling and punctuation, toward meaning-level revisions involving clause structure, cohesion, and argument development. The chatbot's explicit, metalinguistic feedback encouraged learners to notice recurring error patterns and apply self-correction in subsequent drafts. Learners' dependence on external correction gradually decreased as their confidence in self-editing increased. These findings support corrective feedback theory, the noticing hypothesis, and computer-mediated writing feedback frameworks, which emphasize that written accuracy develops through repeated, learner-engaged feedback cycles. This study contributes a new perspective by positioning the AI chatbot not as an automatic error-correction tool, but as an interactive feedback partner that scaffolds learners' revision competence. Chatbot-mediated feedback can therefore serve as a practical alternative for sustained writing improvement in contexts where individualized teacher feedback is limited.