Acceso usuarios
[1]
;
Januardi, Josephine I.
[1]
;
Bonar, Gary
[2]
;
Kurniawan, Daniel
[1]
Indonesia
Australia
Research on AI-powered feedback in language learning increasingly focuses on generative tools, yet automated written corrective feedback (AWCF) technologies such as Grammarly remain essential for fostering autonomous grammar learning and supporting meaning-making in L2 writing. As lower-proficiency learners face greater challenges when using AWCF, this study examines how EFL learners at A2-B2 CEFR levels interact with Grammarly feedback. Using a mixed-method case study design, the research explores learners’ behavioural, cognitive, and affective engagement. B1-B2 learners engaged more selectively and reflectively, demonstrating stronger self-regulation, whereas A2 learners were more reactive, surface-focused, and emotionally vulnerable. Yet engagement also varied within A2 levels, indicating that proficiency alone did not determine learners’ strategies in using Grammarly. This research advanced AWCF research by showing how proficiency, metalinguistic knowledge, trust, confidence, critical stance, and digital experience interact dynamically to shape engagement, and how trust could function both as a resource and a risk, shaping confidence, decision making, and the depth of cognitive engagement. Across the group, Grammarly was mainly employed to correct surface errors rather than to promote in-depth learning. These findings highlight the limitations of problem-solving technologies for learners with fewer linguistic resources and call for process-oriented instruction that builds interpretive and reflective capacities to foster deeper, more autonomous engagement with AWCF.
