Effects of Voice‐Based ChatGPT on Chinese‐as‐a‐Second‐Language Learners’ Mandarin Pronunciation

ABSTRACT This study investigates the impact of voice‐based ChatGPT on the Mandarin pronunciation of Chinese‐as‐a‐second‐language (CSL) learners. Learners’ oral productions, collected as they read aloud their self‐written stories, were analyzed to determine the extent of pronunciation improvement and to explore their perceptions of the tool. The participants consisted of 31 university students from Vietnam, Japan, and Thailand. Data included four audio recordings of student readings and a questionnaire survey. The study addressed two research questions: (1) How do CSL learners’ Mandarin pronunciation skills change over time through interaction with voice‐based ChatGPT? and (2) What are CSL learners’ perceptions of using voice‐based ChatGPT for pronunciation practice? Results indicated consistent improvement across the recordings, with the most significant gain observed between the first and fourth recordings. In terms of specific dimensions assessed, overall comprehensibility (the listener's perceived ease of understanding the speaker's utterance) improved the most, followed by tone, while accuracy (e.g., initials and finals) showed the least improvement, though all mean differences were statistically significant ( p < 0.05). Regarding learner perceptions, participants reported positive attitudes, perceived usefulness, ease of use, and favorable affective responses (e.g., feeling comfortable). Ratings for all perceptual dimensions were significantly correlated ( p < 0.05). Pedagogical implications of the findings are also discussed in the paper.