Generative AI in L2 English Academic Writing: University Students’ Perspectives on Autonomy and Human–AI Interaction

Generative Artificial Intelligence (GenAI) tools are widely reported to improve writing quality and efficiency, yet their implications for autonomy, defined here as volitional, informed, and self-endorsed decision-making, remain underexplored. Drawing on an integrated Self-Determination Theory (SDT) and Sociocultural Theory (SCT) framework, this study interviewed twenty second-language (L2) English learners in Hong Kong universities about how they perceive and report their interaction practices when using GenAI in L2 academic writing. Hybrid inductive/deductive thematic analysis of the interview data yielded six themes: selective and strategic deployment of GenAI; concerns about dependency; preservation of authorial voice; perceived benefits; critical evaluation of outputs; and awareness of institutional expectations. Learners described balancing efficiency against authentic authorship, editing or reframing AI-generated text to protect their voice, and treating GenAI as a peripheral rather than primary agent. From these reported practices we derived a heuristic of seven roles that learners ascribed to GenAI: text analyst, consultant, literature search/review instrument, tutor, creative aid, adversary, and conversational partner. Implications are drawn for pedagogy that treats GenAI as a mediational tool whose contributions must be subordinated to the writer’s own intellectual agenda.