<![CDATA[This systematic literature review investigates how artificial intelligence transforms speaking practice in English as a Foreign Language (EFL) and English as a Second Language (ESL) contexts through an applied communication framework. Unlike prior reviews that evaluate AI-mediated speaking primarily in terms of linguistic outcomes, this study employs communicative competence theory, the willingness-to-communicate (WTC) model, and computer-mediated communication (CMC) theory as an integrated analytical lens to interpret how AI reconfigures the communicative conditions of oral language development. Following PRISMA 2020 guidelines, the review synthesises 37 peer-reviewed articles published between 2021 and 2026, sourced from Scopus, Web of Science, ERIC, and ScienceDirect. Thematic synthesis generated four analytical categories: AI as a conversational partner, AI as a feedback mechanism, AI as affective support, and AI as a constrained communication environment. Findings indicate that AI-supported practice consistently enhances pronunciation, fluency, and willingness to communicate, whilst also revealing persistent limitations in pragmatic sensitivity, authentic interactional transfer, and equitable access. The principal novel contribution of this review is the conceptual framework of ‘Mediated Communicative Scaffolding’, which repositions AI not as an automated correction tool but as a structured communicative environment that prepares learners for authentic human interaction. This theoretically grounded framework offers practical guidance for integrating AI into communication-oriented speaking pedagogy.]]>
