Hesham Aldamen, Mohamad Almashour, Mutasim Al‐Deaibes, Marwan Jarrah, Rami Alsharefeen
This mixed-methods quasi-experimental study examined whether AI-mediated role-play could support complaint-handling performance in an English for Tourism and Hospitality course. The study was conducted with 89 intermediate-level undergraduate EFL students at a public university in Jordan. Students were assigned by class section to a control group that completed instructor-led role-play and an experimental group that engaged in AI-mediated role-play supported by structured AI feedback. Over a six-week intervention, both groups practiced customer complaint scenarios targeting politeness, language accuracy, and problem solving. Quantitative data were drawn from pre-test and post-test performance tasks scored with an analytic rubric, and qualitative data came from AI feedback records, instructor feedback, student interviews, and reflection entries analyzed through thematic analysis. Both groups improved from pre-test to post-test, but the experimental group showed larger gains across all rubric dimensions. The qualitative findings showed that students valued the immediacy, organization, and repeatability of AI feedback, especially during revision, while continuing to rely on instructor feedback for contextual nuance, pragmatic judgment, and strategy guidance. The study contributes to current discussions of instructional AI in applied linguistics by offering an ESP-focused account of how structured AI role-play can extend practice opportunities when it is carefully aligned with course outcomes, guided by the instructor, and framed through responsible use.