The impact of generative AI-powered chatbots on L2 comprehensibility

Abstract While generative AI-based chatbots expand opportunities for L2 pronunciation practice, not all are designed for language learning or provide explicit feedback. Through a comparison of two chatbots, Pronounce , which offers explicit pronunciation feedback, and Gemini , a general-purpose chatbot whose real-time transcription may serve as implicit feedback, this study explored whether practice with these chatbots had an impact on L2 English learners’ comprehensibility and whether any improvements were influenced by the presence of explicit feedback. Three groups of learners participated: two experimental groups, each practicing with one of the chatbots, and one control group. Although comprehensibility ratings indicated no statistically significant improvements at the group level based on training or the specific chatbot used, individual learners demonstrated improvements. These advancements were noted among motivated learners who completed most of their speaking sessions. Learners had positive impressions of their experience with the chatbots and believed that their practice contributed to their pronunciation improvement.