This study investigates the effectiveness of voice-based AI chatbots, such as C-AI, Parlalingo, and Copilot
in reducing foreign language classroom anxiety among English Language Education students. Speaking
anxiety remains a major barrier to students’ willingness to communicate, particularly in EFL contexts
where opportunities for authentic interaction are limited and fear of negative evaluation is high. Hence,
this research aims to examine whether incorporating voice-based AI chatbots over a ten-week period can
significantly reduce students’ anxiety levels. A quantitative pre-experimental design was employed
involving 33 students. Before conducting pre-test and post-test, data were collected using an adapted
version of the Foreign Language Classroom Anxiety Scale (FLCAS). Descriptive statistics, Shapiro–Wilk
normality test, and paired-samples t-test were used to analyse the data. The results showed that the mean
anxiety score declined from 3.31 (SD = 0.37) in the pre-test to 3.03 (SD = 0.38) in the post-test. The pairedsamples t-test showed that this decrease was statistically significant (p < .001) with a moderate-to-large
effect size (Cohen’s d = 0.74). However, challenges such as occasional technical problems and the need
to verify AI-generated responses were also described. Overall, the findings suggest that voice-based AI
chatbots have the potential to reduce speaking anxiety and foster positive learning experiences. This study
highlights the promising role of voice-based AI chatbots in supporting both affective and pedagogical
aspects of language learning. Further study employing more rigorous experimental designs is
recommended to establish their effectiveness more conclusively.