Authentic and functional dialogue with AI ‐based student avatars: Evidence from a practice‐based teacher education program

Abstract Interactions with AI‐based student avatars offer scalable opportunities to practice educational dialogue in practice‐based teacher education (PBTE), yet little is known about the interaction quality of such simulations. This exploratory case study examined whether avatar‐mediated exchanges provide authentic and functional learning opportunities for teacher education. N = 21 preservice teachers completed a 10‐minute, one‐to‐one conversation with an AI student avatar embedded in a PBTE module on leading educational dialogue. The study combined post‐session questionnaires, system metadata, transcript coding and interviews. Participants perceived the exchanges as clearly functional and moderately authentic. Metadata indicated balanced participation, with slightly longer teacher turns. Transcript analyses likewise pointed to high functional dialogue alongside moderate authentic dialogue. Further speaker‐differentiated analyses showed the avatar acting as a responsive counter‐voice, while preservice teachers applied dialogic talk moves. Interview data supported these findings, with participants describing the avatar's language and stance as plausible and highlighting the avatar's contingent uptake of prior turns, while also noting limits to full classroom realism. Taken together, the findings suggest that avatar‐mediated dialogues can provide a meaningful learning opportunity for novices in PBTE. Implications for the design and educational use of AI‐based avatars in teacher education are discussed. Practitioner notes What is already known about this topic Practice‐based teacher education (PBTE) uses approximations of practice to help preservice teachers rehearse core teaching practices. Dialogic pedagogy (eg, pressing for reasons, building on ideas and taking up students' contributions) supports deeper understanding and participation. AI‐driven avatars are emerging as low‐cost, scalable tools for simulated practice, but evidence on the quality of avatar–teacher interaction is still limited. What this paper adds Implementation‐based evidence that short, one‐to‐one conversations with an AI student avatar are experienced as clearly functional and moderately authentic by preservice teachers. Mixed data (self‐reports, system metadata, transcript coding, interview excerpts) show balanced turn‐taking and reveal the avatar's role as a responsive ‘counter‐voice’, while preservice teachers applied dialogic talk moves to develop ideas, elicit reasoning, and challenge claims. A transparent description of the scenario design and prompts that educators can replicate or adapt. Implications for practice and/or policy AI avatars can provide a manageable and accessible opportunity to practice educational dialogue before moving to more complex classroom interactions. Scenario design matters: controversy‐ready prompts and contingent uptake may support explanation, justification and dialogic engagement in avatar‐mediated rehearsals. PBTE programs should combine avatar‐based rehearsal with guided debriefing and opportunities for repeated cycles of rehearsal and reflection.