<![CDATA[Background: Generative AI has rapidly entered EFL academic writing, yet its contribution remains contested because polished output may obscure limited revision reasoning, weakened authorship, and emerging dependence on automated feedback. Objective: This study examines how generative AI assistance shapes revision quality, feedback uptake, writer development, and dependency risk among Indonesian EFL students. Method: Using qualitative public-document corpus analysis, this study analysed 18 verified sources published between 2023 and 2026, comprising Indonesian empirical studies, international comparative research, and institutional or policy documents on AI-assisted writing. Results: Findings show that AI assistance is most frequently positioned as a revision and feedback mechanism, particularly for grammar, vocabulary, coherence, organization, and essay quality. Findings also indicate that AI feedback becomes pedagogically developmental only when students evaluate, modify, reject, or justify suggestions instead of accepting machine-generated revisions as finished text. Implication: Findings further reveal that teacher mediation, disclosure rules, assessment design, academic integrity guidance, and critical AI literacy are necessary to prevent AI-supported revision from becoming cognitive offloading. Novelty: This study offers a mediated model of AI-assisted EFL writing that links revision quality, writer agency, dependency risk, and pedagogical governance within Indonesian higher education.]]>
Copyrights © 2026
