AI in EFL blended learning: impact on speaking performance and learning resilience

Abstract This study investigates the impact of AI-assisted blended instruction on EFL learners’ speaking performance and learning resilience. Grounded in Social Cognitive Theory, the Technology Acceptance Model, and Sociocultural Theory, the research addresses a growing need to understand how AI tools shape not only linguistic accuracy but also learners’ psychological adaptability. Adopting a quasi-experimental design, the study compared an experimental group using ChatGPT and Gemini within a task-based learning framework to a control group receiving conventional instruction over an 18-week period. Quantitative analyses revealed significant gains in pronunciation and accuracy among the experimental group, with coherence showing marginal improvement and fluency remaining unchanged. In terms of resilience, learners demonstrated marked increases in metacognitive and social resilience, whereas ego resilience remained unaffected. These findings suggest that while AI tools can effectively support linguistic development and foster strategic learning behaviors, they may be less effective in cultivating emotional adaptability. The differentiated outcomes highlight the importance of aligning specific AI functionalities with targeted pedagogical goals rather than assuming uniform benefits. This study contributes a nuanced perspective to the discourse on AI in language education by uncovering domain-specific effects and clarifying the boundaries of AI’s pedagogical influence. It calls for future research to explore the long-term impacts of AI integration on learner autonomy, motivation, and affective growth, thereby paving the way for more intentional, theory-driven applications of AI in EFL contexts.