AI Scaffolding and the Development of a Clinical Voice: International Medical Students in an ESP Context

Background As generative AI reshapes medical education, its role in supporting linguistically diverse learners engaged in high-stakes clinical communication remains underexplored. Building on prior research showing that traditional interactive formats (e.g., group presentations) foster authentic dialogue but impose heavy preparation burdens, this study examines how international medical students use AI as pedagogical scaffolding during simulated Multidisciplinary Team Meetings (sMDTMs) in an English for Specific Purposes (ESP) context. It explores how students negotiate AI’s affordances to develop an authentic clinical voice, addressing both its benefits and emerging challenges. Methods Using a qualitative, scaffolding-informed framework, 42 international medical students at an Iranian university employed ChatGPT, Gemini, or Claude to prepare for sMDTMs. Data were triangulated across student reflective journals, peer evaluations, and expert assessments of video-recorded sessions. Reflexive thematic analysis traced how AI-mediated scaffolding influenced learners’ preparation and performance. Results Students viewed AI as a dual-edged scaffold. While most (92.9%) valued its efficiency and language refinement, many (78.6%) critiqued the generic, impersonal quality of AI outputs, prompting active “humanizing” edits through personalization and simplification. Expert reviewers identified a recurring scaffold-transfer gap, where AI-polished fluency masked conceptual weaknesses during spontaneous interactions. Prior English-mediated experience moderated AI use: less experienced learners relied on AI for foundational linguistic scaffolding, whereas others used it for rhetorical refinement. Conclusions Generative AI transforms rather than replaces the social dimensions of clinical communication learning. While it alleviates workload pressures, it also introduces authenticity and dependency challenges. The study advocates a hybrid pedagogical model combining AI’s efficiency with structured peer dialogue and reflective de-scaffolding to foster independent, authentic professional communication.