Developing L2 turn-taking with ChatGPT: A longitudinal conversation analytic study

This conversation analytic study examines how a Korean EFL learner’s turn-taking evolved through 30 role-play sessions with ChatGPT-4 over three months. Adopting a hybrid design that integrates conversation analysis with quantitative metrics, the research tracks systematic changes in turn complexity, diversity, design, and timing across three developmental phases. The findings demonstrate that while sustained interaction with the AI chatbot facilitates significant gains in L2 interactional competence (IC), certain systemic features such as rigid pause thresholds and AI verbosity simultaneously act as interactional constraints. Despite these limitations, the learner demonstrably recategorized ChatGPT as a distinct AI chatbot interlocutor and adaptively recipient-designed her talk accordingly, while ChatGPT functioned as a “competent-enough” L2 conversational partner capable of scaffolding L2 IC development. Finally, this research provides pedagogical insights for integrating AI-mediated practice with human-mediated, multimodal instruction.