Generative AI Literacies and Discipline-Specific Knowledge: Preservice Teachers Planning Task-Based L2 Lessons with GenAI

Abstract This study investigates the use of Large Language Models (LLMs) to support Task-Based Language Teaching (TBLT)-oriented lesson planning in a preservice teacher education context, examining the extent to which LLM use can contribute to the mobilisation of TBLT-related competencies. The case study draws on triangulated data from surveys, interaction logs with the LLM, lesson plans produced by preservice teachers and their reflections. The module required participants to generate a lesson plan with an LLM and to compare the resulting plan with a previously created lesson plan developed without AI support. The analysis focuses on knowledge of TBLT principles and their implementation during lesson planning, and the GenAI competence activated through the interaction with the LLM. The findings reveal considerable variation in how the dyads interacted with the LLM. Some dyads formulated more precise prompts and prompt iterations, provided more contextualised input, engaged more critically with the LLM-generated output, and iteratively refined their lesson plans in ways that appeared more consistent with TBLT principles. The data suggest an association between a stronger conceptual understanding of TBLT and more purposeful interaction with the LLM. This association indicates that disciplinary knowledge may facilitate the effective use of LLMs for TBLT-oriented lesson planning. Rather than replacing pedagogical expertise, LLM use appears to be most productive when preservice teachers are able to critically evaluate and adapt the generated output in light of established TBLT principles.