From sporadic to sustained: facilitating AI-mediated informal digital learning of English (AI-IDLE) speaking practice through a teacher support framework

Purpose While generative AI empowers language learning beyond the classroom, AI-mediated informal digital learning of English (AI-IDLE) speaking practice tends to be aimless and sporadic. Drawing on Activity Theory and the Social Support Framework, this study proposed and empirically validated a teacher support framework designed to resolve the systemic contradictions inherent in self-directed AI-IDLE speaking practice.Methodology Employing a mixed-methods quasi-experimental design over an 18-week intervention, the study involved 108 EFL undergraduates across three conditions: baseline IDLE using non-AI digital resources (CG, n = 34), self-directed AI-IDLE using the AI chatbot Doubao (EG1, n = 32), and teacher-supported AI-IDLE using the same tool (EG2, n = 42). Data on speaking proficiency, speaking self-efficacy, and speaking anxiety were collected via pre- and post-tests, triangulated with semi-structured interviews from 22 participants.Findings ANCOVA results confirmed the framework’s effectiveness, revealing a significant proficiency hierarchy (EG2 > EG1 > CG) in fluency, vocabulary, and grammar, alongside superior gains in speaking self-efficacy and reduced speaking anxiety for EG2. No significant differences were found in pronunciation. Qualitative analysis revealed that the teacher’s’ instrumental, informational, emotional, and appraisal support functioned synergistically to resolve technological and affective tensions.Originality The findings contribute to the understanding of the teacher’s role in the AI era, demonstrating how educators can shift from knowledge transmitters to designers of self-directed learning environments that maximize the pedagogical potential of generative AI.