oai:eurokd.com:article/2170 2026-06-29

AI-Mediated Written Corrective Feedback in L2 Writing: What Changes, What Doesn’t, and What We Still Don’t Know Volume 4 Filomachi Spathopoulou Konstantinos M. Pitychoutis Written Corrective Feedback Large Language Models (LLMs) Complexity-Accuracy-Fluency (CAF) Feedback Literacy Second-Language Writing Assessment Integrity <p style="text-align: justify;">This reflective commentary revisits key insights from research on written corrective feedback (WCF) within the context of feedback facilitated by large language models (LLMs) in higher-education L2 writing. The CAF framework shows that accuracy is most consistently improved through targeted, well-designed feedback. In contrast, claims about complexity and fluency should be approached cautiously, as they depend on factors like task design, sequencing, and learner engagement. The paper distinguishes between traditional automated writing assessments and the generative, dialogic features of LLMs. It proposes a straightforward approach for teachers to provide feedback, including specific prompts, metalinguistic explanations, learner reformulation, genre confirmation, and reflection periods. The paper emphasises that the educational value of LLM-based WCF relies more on thoughtful mediation, ethical considerations, and alignment with assessment goals than on speed or novelty. It concludes by highlighting priority areas for future research, such as the durability of learning gains, transfer from assisted to independent writing, the interaction of dose, timing, and scope, maintaining learner engagement, and ensuring equitable access to AI-mediated feedback across proficiency levels.</p> Feedback Research in Second Language 2026-06-29 Text https://api.eurokd.com/Uploads/Article/2170/frsl.2026.04.02.pdf https://doi.org/10.32038/frsl.2026.04.02 en Pages 6–15