This study was caried out to determine the extent to which authentic videos within a CLIL framework improve 4th-year medical students’ English communication with patients and to explore students’ perceptions of this approach. An action research was designed at Thai Nguyen University of Medicine and Pharmacy among 55 fourth-year students completed a CLIL intervention that integrated authentic doctor–patient videos into EMP lessons. Outcomes were measured with OSCE-style communication tasks (pre/post) and a perception questionnaire (5-point Likert scale). The results of the study shows that students’ OSCE communication scores improved significantly from pre-test (M = 26.65, SD = 4.99) to post-test (M = 31.35, SD = 3.86), t(54) = −17.41, p < .001; mean gain = 4.69 points. Pre–post scores were strongly correlated (r = .93, p < .001). Perception data indicated high engagement and perceived value of authenticity (e.g., visually engaging videos M up to 5.00; real-life scenarios M = 4.89), with reported gains in appropriate medical language use (M = 4.59), vocabulary (M = 4.56), cultural/pragmatic awareness (M = 4.85), and pronunciation/intonation (M = 4.27). Speaking confidence improved moderately (M = 3.80). Learner autonomy indicators were mixed, suggesting a need for additional scaffolding. It is concluded that authentic videos embedded in CLIL significantly enhanced students’ English-medium clinical communication and were positively received, particularly for realism, linguistic development, and pragmatic competence. To maximize impact, video-based input should be paired with structured speaking practice and autonomy-supportive tasks. Findings inform EMP curriculum design, teacher training, and assessment reform toward performance-based evaluation.