The Impact of Generative AI on Learner Autonomy and Critical Thinking in English as a Foreign Language (Efl) Writing Classrooms

The rapid integration of generative artificial intelligence (AI), particularly large language models (LLMs) such as ChatGPT, has reshaped pedagogical practices in English as a Foreign Language (EFL) writing classrooms. This article synthesizes empirical findings published between 2023 and 2025 to investigate how such tools influence two interrelated dimensions of language learning: learner autonomy and critical thinking (CT). A structured review of peer-reviewed studies including quasi-experimental interventions, mixed-methods research, and systematic reviews was conducted. Evidence consistently shows that AI-mediated writing instruction enhances writing performance through immediate, context-sensitive feedback, improved drafting strategies, and iterative revision processes. For example, studies with n=136 university-level EFL students demonstrated statistically significant improvements in content, organization, and mechanics after AI-supported interventions, while another mixed-methods study involving 70 experimental and 66 control group participants reported large within-group effects favoring AI integration. Similarly, a one-group pre-post study of n=117 students found highly significant improvements (p < .001) across multiple dimensions of ESP writing. Beyond performance, evidence indicates positive shifts in learner autonomy, with students demonstrating increased responsibility for drafting, evaluating AI-generated feedback, and monitoring their revision decisions. However, findings related to critical thinking are more nuanced. A 2025 systematic review reported that approximately two-thirds of studies (≈67%) found enhanced CT outcomes, including improvements in argument quality, reasoning, and evaluation, whereas about one-third (≈33%) cautioned that uncritical reliance on AI may suppress independent analysis and reflective judgment. These outcomes suggest that task design and pedagogical scaffolding are decisive in shaping whether AI use strengthens or weakens higher-order thinking. The analysis is interpreted through Self-Determination Theory (Ryan sociocultural theory, positioning LLMs as mediational tools in learners’ zones of proximal development; and feedback-as-dialogue frameworks, which reconceptualize AI outputs as opportunities for feedback literacy and metacognitive engagement. Drawing on this evidence, we propose a pedagogical design pattern, AI-scaffolded, teacher-orchestrated, rubric-anchored writing that integrates generative AI as a supportive resource while ensuring learners remain critically engaged and responsible for their own learning processes. Overall, the findings confirm that generative AI has the potential to serve as a catalyst for improved writing quality, greater autonomy, and the development of critical thinking in EFL classrooms, but only when embedded within intentional, reflective, and ethically guided instructional practices.