Audio-Based Discussions in Blended Learning: A Qualitative Study of Chinese University Students' English-Speaking Skills

English speaking skills are crucial for effective global communication. However, in EFL contexts such as China, where an exam-oriented education system prevails, the focus on developing oral skills is often neglected. Although blended learning has successfully enhanced English speaking proficiency, audio-based discussions, which provide a platform to balance speaking authenticity with social pressure, have received relatively little attention. This research aims to (1) explore how university students perceive their experiences with audio-based discussions, (2) examine the influence of audio-based discussions on their speaking anxiety, and (3) investigate the impact of audio-based discussions on specific speaking skills involving pronunciation, fluency, vocabulary, and grammar. The qualitative research involved semi-structured interviews with 11 first-year students at a selected Chinese university. The results show that students perceived audio-based discussions as a way to enhance engagement and promote genuine, reflective communication. Furthermore, initial anxieties among students regarding pronunciation and unfamiliarity with the technology gradually diminished with consistent practice and the presence of a virtual audience. Significant improvements were observed in pronunciation, fluency, and vocabulary, but students reported minimal progress in grammar. These findings demonstrate that audio-based discussions are a valuable and effective platform for addressing challenges in EFL speaking. This study highlights the practicability of incorporating audio-based discussions into blended learning environments and advocates for the integration of AI tools to provide targeted grammatical feedback during speaking activities.