The Impact of GAI-Based Virtual Speaking Partner on Chinese EFL Learners' Speaking Self-Efficacy: An Empirical Study

In Chinese English as a Foreign Language (EFL) education, traditional instruction has overemphasized grammar and writing, limiting the development of learners’ speaking proficiency and causing anxiety and low confidence. The emergence of Generative Artificial Intelligence (GAI) provides new opportunities to address these challenges. This study examines the effects of a GAI-based virtual speaking partner on Chinese EFL learners’ speaking self-efficacy and their user experiences. A mixed-methods design was employed with 40 undergraduate students randomly assigned to either an experimental group or a control group. The experimental group engaged in speaking practice with the GAI-based virtual partner, whereas the control group followed traditional speaking practice. Data were collected through a speaking self-efficacy scale and semi-structured interviews. Results indicate that, after an eight- week intervention, the experimental group significantly outperformed the control group. Learners valued immediate feedback, reduced speaking anxiety, and enhanced autonomy in practice, but noted limitations in emotional responsiveness, speech recognition, and conversational depth. These findings suggest that GAI-based virtual speaking partner can serve as effective supplementary tools in speaking English instruction and provide empirical evidence for the integration of intelligent technologies into language learning.