Catalysts, Not Only Constraints: Pre-Service EFL Teacher Agency in Generative AI-Assisted Lesson Planning

This study explores how pre-service English as a foreign language (EFL) teachers used three Generative AI (GenAI) tools, ChatGPT, Gemini, and DeepSeek, while planning and teaching lessons. Drawing on ecological model of teacher agency, it examines how perceived limitations of these tools were associated with different forms of teacher agency, and whether the limitations associated with each form varied across tools. Using a qualitative case study design, the study draws on 58 participants’ reflective essays and observations of 13 classroom lessons. Findings reveal three forms of agency: recontextualizing (adapting generic AI content locally), substitutive (rejecting and replacing impractical suggestions), and improvisational (spontaneous in-class adjustments), each associated with different types of limitations. As a conceptual contribution, the study proposes a three-part typology of teacher agency based on the chordal triad model, and offers recommendations for EFL teacher education, highlighting the need for critical AI literacy and context-sensitive use of multiple GenAI tools.