<![CDATA[Speaking anxiety and limited opportunities for authentic oral practice are among the most persistent barriers to spoken English development in EFL microteaching programmes. This study examines whether conversational AI chatbots can function as low-stakes speaking partners that reduce anxiety and improve oral performance readiness for pre-service EFL teachers before microteaching delivery. Using a quasi-experimental design with pre- and post-test oral performance assessments, Foreign Language Speaking Anxiety Scale scores, and qualitative data from reflective journals and post-study interviews, the study compared two groups of pre-service EFL teachers at a public Islamic university in Yogyakarta, Indonesia: one group engaging in structured chatbot-mediated oral practice sessions twice weekly for eight weeks before microteaching, and a control group receiving conventional preparation support only. Findings indicate significantly reduced speaking anxiety and improved oral fluency, pronunciation accuracy, and interactional competence scores in the chatbot group, with qualitative data revealing that the low-stakes, non-judgemental character of chatbot interaction was the most consistently valued feature—particularly for participants with high baseline anxiety. However, the study also identifies a transfer gap: the communicative behaviours practised with chatbots did not automatically transfer to human-audience microteaching contexts, and several participants reported that the emotional experience of a chatbot interaction and a live teaching performance remained categorically different. The study draws on Willingness to Communicate theory (MacIntyre et al., 2024) and Vygotskian scaffolding frameworks to argue that chatbot-mediated oral practice is a valuable but insufficient preparation for microteaching—most effective when embedded in a broader pedagogical sequence that explicitly bridges from low-stakes AI interaction to high-stakes human performance.]]>
Copyrights © 2026
