FOSTERING LEARNER AUTONOMY IN EFL WRITING THROUGH ARTIFICIAL INTELLIGENCE: A SYSTEMATIC REVIEW

Abstract

Background: Learner autonomy has been a key focus in L2 writing instruction for many years. In recent times, the use of AI writing tools has further heightened the interest in how these tools assist in learner autonomy when engaging with writing. The empirical literature is not yet consolidated, fragmenting the tools, constructs and designs. Purpose: This review combines evidence (2021-2026) to understand the role of AI in EFL learners' autonomy in writing, shaping around five research questions concerning the use of tools/designs, the mechanism, the outcomes, the moderators, and gaps in evidence. Methods: After the PRISMA 2020 process, 20 primary studies were identified. A narrative-thematic synthesis was carried out and appraised for quality using the MMAT, due to the design and outcome heterogeneity. Results: Most studies indicated that AI-mediated instruction, such as AWE, generative AI, and developed systems, is linked to increased autonomy, self-efficacy, SRL strategy use, student engagement, and writing performance. Six themes were formulated: (1) the construction of autonomy in a feedback system, (2) SRL scaffolding, (3) affective/motivational outcomes, (4) writing performance, (5) countervailing evidence (overreliance risk, diminishing returns for highly self-regulated learners). The effects were mitigated by proficiency, education level, tool complexity, duration, and human–AI integration. Conclusion: The potential of AI tools to enhance EFL writing autonomy is evident yet not uniformly distributed across different functions, most notably in "humans-in-the-loop" models, but not to the exclusion of teacher scaffolding. Pedagogy, policy and teacher development implications are considered, as well as a research agenda focused on standardized assessments and longitudinal studies.