“Better safe than sorry”: Unveiling doctoral students’ critical AI literacy development in academic writing

Despite the growing incorporation of GenAI tools in academic writing, research has yet to adequately examine how doctoral students develop critical AI literacy in their interactions with such technologies. Adopting a sociocultural perspective, this qualitative case study investigates the processes through which nine doctoral students developed critical AI literacy in academic English writing. Drawing on semi-structured interviews, screen recordings, learning journals, reflective reports and AI logs, data analysis reveals four developmental trajectories of critical AI literacy: from uncritical consumers to informed users who examined GenAI tools' inner workings and limitations; from AI-dominated users to dominators of AI who valued text ownership and authorial agency; from passive recipients to strategic users who emphasized proactive GenAI use for self-learning; and from efficiency-oriented users to responsible scholars attentive to GenAI's epistemic, ethical and social implications. These changes were mediated by their interactions with knowledgeable others, engagement with cultural artefacts, and self-regulated exploration. The study highlights the socially-mediated process of critical AI literacy and offers practical implications for cultivating critical AI literacy in doctoral education.