Effects of AI-assisted speaking practice on spoken English performance in higher vocational education

Abstract

Abstract With the rapid advancement of artificial intelligence (AI), AI-assisted tools are increasingly integrated into language education. However, empirical evidence on their effectiveness for spoken English in higher vocational colleges remains limited. This quasi-experimental study examined whether AI-assisted speaking practice using iFLYTEK Spark, a generative AI platform with speech recognition and real-time feedback capabilities, was associated with higher post-intervention spoken English performance among 92 first-year vocational college students in China. Two intact classes were assigned to an experimental group ( n = 46) receiving 12 weeks of AI-assisted practice plus conventional instruction and a control group ( n = 46) receiving conventional instruction only. Post-intervention speaking performance was assessed by three trained raters using an analytic rubric covering pronunciation, language use, fluency and coherence, and interaction. The experimental group achieved significantly higher overall posttest scores (M = 3.69, SD = 0.76) than the control group (M = 3.19, SD = 0.82; t(90) = 3.011, p = 0.003, Hedges’ g = 0.63). Medium-to-large effect sizes were found in language use (g = 0.49), fluency and coherence (g = 0.50), and interaction (g = 0.75), whereas the difference in pronunciation was smaller and non-significant (g = 0.38, p = 0.071). Semi-structured interviews with one teacher and three students provided supplementary insights into perceived benefits, including increased practice opportunities, reduced speaking anxiety, and enhanced motivation. While the intact-class design and coarse baseline proficiency measure limit causal claims, the findings suggest that AI-assisted tools can serve as a valuable supplement to conventional spoken English instruction in higher vocational settings. Implications for pedagogy and future research directions are discussed.