<![CDATA[Generative artificial intelligence is increasingly used in English as a Foreign Language (EFL) instruction, but evidence from classroom-randomized studies remains limited. This study evaluated an eight-week teacher-supervised ChatGPT-assisted EFL program using a matched-pair cluster-randomized controlled trial. Sixteen intact first-year undergraduate classes from two universities in East Java, Indonesia (N = 320) were paired by institution and instructor and randomized within pairs to ChatGPT-assisted instruction (8 classes; n = 160) or comparison instruction without generative AI (8 classes; n = 160). Engagement, self-determined learning motivation, and foreign-language enjoyment were assessed before and after the intervention. Primary analyses used ANCOVA adjusted for corresponding pre-test scores, English proficiency, and matched-pair fixed effects, with class-clustered finite-sample standard errors; exact randomization inference enumerated all 256 allocations permitted by the design. Adjusted mean differences favored ChatGPT-assisted learning for engagement (0.172, 95% CI [0.075, 0.269]), motivation (0.214, 95% CI [0.146, 0.281]), and enjoyment (0.218, 95% CI [0.109, 0.328]). Holm-adjusted exact p values were .047, .023, and .047, respectively. Structured ChatGPT use may enhance students’ affective and behavioral learning experiences when embedded in teacher-guided prompting, verification, and revision. These findings provide empirical support for the responsible integration of generative AI in EFL classrooms and highlight the importance of teacher mediation in ensuring meaningful, reflective, and pedagogically grounded AI-supported learning.]]>
Copyrights © 2026
