<![CDATA[The rapid integration of Artificial Intelligence (AI) in higher education offers efficient writing support but poses a critical risk of learner dependency, which can decease critical thinking and academic integrity. This study aims to investigate the learning strategies employed by Indonesian EFL postgraduate students to plan, monitor, and evaluate their engagement with AI chatbots during academic writing. Employing a qualitative multiple-case study design, data were collected from eight postgraduate students through semi-structured interviews and stimulated recall sessions based on their authentic AI conversation logs. The findings reveal that the students rigorously applied Self-Regulated Learning (SRL) strategies across three interaction phases to safeguard their intellectual independence. During the Planning phase, learners consciously limited the AI's role to preparatory tasks and brainstorming, strictly preventing the machine from generating full drafts. In the Monitoring phase, participants actively filtered and synthesized the generated content, frequently positioning the chatbot as a mock examiner to rigorously test their academic arguments. Furthermore, within the evaluation phase, the students executed multi-layered evaluations by manually verifying AI-generated citations against reputable academic databases to mitigate hallucinations, followed by consultations with human experts’ validation. Ultimately, the study concludes that strong metacognitive regulation effectively prevents cognitive laziness, ensuring that AI serves strictly as an adaptive linguistic model rather than a replacement for human intellect.]]>
Copyrights © 2026
