Postgraduate Students' AI Engagement in L2 Academic Writing: Toward Student‐Led Human– AI Collaborations

Abstract Generative artificial intelligence (AI) is transforming the teaching and learning of second language (L2) writing. Despite AI's affordances in assisting L2 writing, improvements in writing skills and learning outcomes that extend beyond producing polished texts largely depend on students' active AI engagement, typically involving AI‐generated content and AI tools. This study adopts a qualitative, classroom‐based inquiry to examine how 56 L2 postgraduate students engage with AI in an authentic academic writing class. Framing engagement as a multidimensional meta‐construct, we identify a contextualized pattern of engagement characterized by students' agential role in shaping human–AI interactions across behavioral, cognitive, emotional, and social dimensions. Specifically, behavioral engagement occurs at compositional and operational levels, whereas cognitive engagement involves cognitive autonomy and metacognitive awareness. Students' emotional engagement involves appreciation and relief, as well as caution and emotional distance. Social engagement intersects with other dimensions, potentially shaping how students act, think, and feel in relation to AI through collaborative reflection. The findings inform a reconceptualization of engagement as a fluid, situated process, moving beyond a simple engagement–disengagement dichotomy toward a dynamic continuum shaped by the ongoing interplay across its multiple dimensions, encompassing both synergies and tensions. Implications for L2 writing and learning are also discussed.