AI versus teachers and peers: a comparative study on speaking performance, willingness to communicate, self-perceived communicative competence and foreign language anxiety

Abstract While online language education is widely accepted for its convenience, it often leads to a relatively isolated learning experience. In response to positive psychology’s emphasis on learner engagement and well-being, this study investigated how an instructional AI chatbot and an AI-based speaking assessment system, introduced midway through the course, could enhance instruction and support speaking practice. Utilizing three-round speaking tests and questionnaires, supplemented by semi-structured interviews, this study examined how AI-supported language learning influenced learners’ speaking performance, along with key affective variables closely related to the speaking experience: Willingness to Communicate (WTC), Self-Perceived Communicative Competence (SCC), and Foreign Language Anxiety (FLA). A total of 103 Chinese EFL learners participated in the experiment. The quantitative results indicated that AI, compared to human teachers and peers, was more effective in enhancing learners’ speaking performance, WTC, and SCC. The qualitative data suggest that this may be attributed to learners feeling more comfortable and secure when producing English in AI-mediated online environments, which reduced social pressure and fostered greater self-expression. In addition to identifying the limitations of AI-supported speaking instruction, the study offers practical guidance for teachers seeking to incorporate diverse AI tools into online classroom activities to foster both linguistic development and positive emotional experiences.