AI-ENHANCED ENGLISH LANGUAGE LEARNING: EFFECTS ON PROFICIENCY AND COMMUNICATIVE COMPETENCE

Abstract

This review examines what AI-enhanced English language learning appears to change and where the evidence remains less conclusive. Intelligent tutoring, automated feedback, conversational AI, speech technologies, and generative AI are considered as distinct pedagogical tools rather than as a single intervention. The quantitative evidence is encouraging. Xu and Wang’s (2024) meta-analysis of 40 empirical studies involving 3,290 participants and 55 effect sizes reported a large overall effect (Hedges’ g = 0.812). Wu and Yu’s (2024) meta-analysis of 24 randomized studies likewise identified a large overall effect for AI chatbots (ES = 0.964), although effect magnitude varied considerably across outcomes and intervention conditions. Individual studies add further detail. Mizumoto et al. (2024) found a strong correspondence between ChatGPT and human judgements of linguistic accuracy (ρ = .79), while Han and Li (2024) examined the use of AI-generated feedback within teacher-mediated EFL writing instruction involving 102 students and four instructors. These findings provide a strong basis for considering AI as a means of expanding practice, feedback, revision, and individualisation. They offer less decisive evidence concerning durable acquisition, independent proficiency, transfer, and communicative competence in socially situated interaction. Drawing on Hymes (1972) and Canale and Swain (1980), the review distinguishes grammatical development from the sociolinguistic, discourse, and strategic resources involved in communication. AI-supported performance may improve substantially while the broader development of communicative competence remains only partly demonstrated.