Mediating Oral Competence Through Gen-AI in Genre-Based Speaking Instruction for Fractured Narratives

Developing oral narrative skills in English as a Foreign Language (EFL) contexts is challenging. It is often made more difficult by limited opportunities for personalized feedback and the cognitive burden of real-time production. This study investigated the integration of GenAI tools, specifically Gemini and ChatGPT, within a Genre-Based Approach (GBA) to scaffold the speaking skills of 32 tenth-grade students in West Java, Indonesia. This study focuses on the creative genre of "fractured stories," which is a demanding task that requires progressive scaffolding. This study used a qualitative case study design to explore how GenAI tools functioned across the GBA cycle. Data were collected through classroom observations, student-AI chat logs, students’ speaking recordings, and semi-structured interviews. The findings revealed that GenAI tools could improve students’ organizational structure and pronunciation by effectively enacting Vygotskian scaffolding, functioning as a dynamic "More Knowledgeable Other," that progressed from a lexical support during the knowledge-construction stage to a personalized phonological coach in the phase of independent practice. Analysis of speaking recordings showed marked improvement in narrative structure across all proficiency levels, while lower-proficiency students demonstrated significant gains in pronunciation accuracy via multimodal rehearsal loops. The study concluded that when embedded in a rigorous genre-based framework, GenAI transforms from a mere text generator into a robust, affective-aware partner that democratizes access to high-quality oral feedback. Future research should examine long-term retention of skills beyond the intervention period and investigate GenAI integration across diverse oral genres.