<![CDATA[This study examines the emergence and interactional functions of spoken grammar features in AI-mediated peer interactions, using a corpus-based discourse analytic approach among English as a Foreign Language (EFL) students. While artificial intelligence is being integrated more deeply within pedagogical contexts, existing studies have paid limited attention to how AI-mediated peer interactions shape learners’ authentic spoken grammar, particularly in naturally occurring collaborative dialogues among EFL students. Twenty EFL students engaged in AI-mediated peer dialogues, producing a learner corpus of spoken interaction derived from chat-based and video-conferenced communication. Data consisted of transcribed peer dialogues, field notes, and follow-up interviews, which were analyzed using qualitative corpus-based discourse analysis. The analysis showed context-appropriate, sustained spoken interaction and core spoken grammar features, including discourse markers (e.g., well, you know), ellipsis, heads and tails, and lexical bundles. These features served practical purposes of turn-taking, elaboration, mitigation, and even emphasis. Contrary to concerns that AI mediation might constrain spontaneous language use, the findings indicate that AI-supported peer interaction facilitated relatively natural and expressive spoken interaction within the study context. Students reported feeling greater relaxation and expressiveness during AI-mediated tasks, suggesting that peer interactions within digital settings foster authentic communicative competence. ]]>
