Exploring the Integration of AI Chatbots into Task-based Language Teaching among Chinese Primary School Students: Associations with English-speaking Fluency and Anxiety

Abstract

This study investigates the persistent challenges in English speaking instruction in lower-tier cities in China, where students often exhibit low English speaking fluency and heightened anxiety associated with limited speaking practice and a reading-centric teaching approach. While Artificial Intelligence (AI) Chatbots (AICT) have shown promise in higher education, empirical evidence regarding their integration into Task-Based Language Teaching (TBLT) at the primary school level remains scarce. To address this research gap and offer a scalable pedagogical contribution, we implemented an 8-week AICT-supported TBLT program involving 68 fifth-grade students. Using a one-group pretest–posttest design, we measured temporal changes in students’ English-speaking fluency (assessed via verbal productivity, hesitation, repetition, and self-correction) and their self-reported speaking anxiety. Data were analysed using paired-sample t-tests. The results revealed statistically significant improvements (p < 0.001) across all fluency dimensions, with particularly large effect sizes in reducing repetition (d = 0.918) and self-correction (d = 0.782), alongside moderate gains in verbal productivity (d = 0.408) and hesitation (d = 0.532). Additionally, the intervention was associated with a comprehensive decrease in speaking anxiety (p < 0.001), with large effect sizes (d > 0.8) observed for 11 of the 20 items. These findings suggest that integrating AICT as an interactive tool within a TBLT framework can help create a supportive, low-anxiety speaking environment, potentially offering a practical pathway associated with the support of oral English teaching in resource-limited primary school contexts in China.