Critical Autoethnography With Multiple Large Language Models: AI ‐Stimulated Reflexive Practice on Language Teacher Educator Identity

Abstract

This study proposes a methodological approach in which large language models (LLMs) serve as reflective and dialogical partners rather than analysts in critical autoethnographic narrative (CAN) research on language teacher educator identity. During an ongoing debate about the use of generative artificial intelligence (AI) in reflexive qualitative research, this approach occupies a narrow methodological space. LLM outputs serve as stimuli for the researcher's reflexive interpretation of CAN texts, while meaning‐making, salience judgment, and theoretical synthesis remain human practices. In a two‐phase design, three LLMs (ChatGPT 5.2 Thinking, Claude 4.5 Sonnet, and Gemini 3 Thinking) first generated reflective stimuli in the form of probing questions and then produced parallel thematic outputs that served only as material for the researcher's reflexive synthesis, not as completed analyses. Four patterns emerged: consciousness‐raising through dialogical awakening, institutional precarity and resistance, pedagogical transformation, and the relational construction of critical identity. The LLMs differed in orientation in ways consistent with poststructuralist epistemology. The contribution is primarily methodological: a procedure for AI‐stimulated reflexive practice in language teacher education research, alongside a critical engagement with the ethical, environmental, and epistemic costs such practice entails.