AI-mediated speaking practice in less-resourced EFL contexts

Effects on fluency and vocabulary

Abstract

AI-based automation is reshaping the way people learn, work, and communicate. This study investigates the effectiveness of AI tools (ChatGPT, Call Annie, Grok, and Gemini) in developing English as a foreign language speaking skills, such as fluency and vocabulary, among learners in less-resourced contexts. The study also identifies participants’ perceptions of the challenges and advantages faced during IELTS preparation. Guided by Vygotsky’s sociocultural theory, the present research used a mixed-methods quasi-experimental approach. The study included two groups of 30 participants each: the experimental group engaged in AI-assisted speaking through one-to-one voice sessions three times a week for 12 weeks, while the control group practiced traditional speaking methods. We collected data through IELTS-formatted pre- and post-speaking assessments and semi-structured interviews. The findings revealed that the experimental group experienced moderate but statistically significant improvements in speech rhythm and lexical resource enrichment. Participants also perceived AI tools as accessible, low-anxiety speaking partners. The findings have implications for equitable language education in resource-limited contexts, as well as for the UNESCO 2030 agenda for inclusive digital learning.