Reflective Regulation of AI-Assisted Translation: Longitudinal Development of Technology Competence and Critical AI Literacy in an Indonesian EFL Classroom

Artificial intelligence (AI) is now routinely used in translation education, but frequent use does not demonstrate critical or professionally responsible technology competence. This study examined how 20 undergraduate EFL translation students regulated AI assistance during a one-semester course. A qualitative-dominant embedded mixed-methods design followed six Indonesian-to-English translation phases structured by Assessment-as-Learning and repeated Plan–Monitor–Evaluate reflection. The dataset comprised 120 reflective journals, phase-based Likert responses used only as descriptive corroboration, and participant-observation notes. Longitudinal thematic analysis began with inductive coding and then interpreted cross-phase patterns through the European Master’s in Translation Technology Competence framework and metacognitive regulation. In Phase 1, 11 students articulated no AI-use plan, six offered vague plans, and three described relatively explicit plans; AI was used mainly for post-draft correction. In Phases 2–3, students increasingly compared, modified, or rejected suggestions according to meaning, grammatical accuracy, academic register, and contextual fit. In Phases 4–6, reflections more consistently defined human authorship and accountability and positioned AI as a verification resource rather than a primary text generator. The trajectory was cohort-dominant rather than uniform, and the design does not establish that reflection alone caused the change. The findings conceptualize technology competence as the reflective regulation of language-mediated human–AI interaction and identify structured metacognitive reflection as a plausible way to develop critical AI literacy in translation education.