A meta-analysis of generative AI effects on language proficiency and affective–cognitive outcomes in language learning

Abstract Generative Artificial Intelligence (GenAI) tools are increasingly integrated into language education, yet their overall effectiveness and the factors shaping their impact remain unclear. This meta-analysis synthesizes evidence from 51 studies reporting 175 independent effect sizes to evaluate the impact of GenAI tools on second and foreign language learning outcomes. Results indicate that GenAI tools produce large, statistically significant effects on language proficiency and affective–cognitive outcomes. Moderator analyses revealed that effects vary by intervention duration, instructional setting, learner characteristics, and target language. Stronger effects were generally observed in informal learning environments, for productive skills, and in studies addressing less commonly taught languages. These findings support sociocultural and interactionist theories of second language acquisition and highlight the need for tailored, context-sensitive integration of GenAI tools. The review offers practical guidance for educators, designers, and policymakers, and underscores the importance of continued research across diverse contexts and learner populations.