Virtual reality (VR) has emerged as a promising technology for training oral communication skills (OCS), resulting in a rapidly growing body of research. We conducted a systematic review of 57 studies (2013–2025) mapping OCS types, communication contexts, and key variables, and introduce a conceptual model to guide future research and practice. Findings reveal that current VR-based OCS training captures only a small part of oral communication. Expanding the Cognitive Affective Model of Immersive Learning (CAMIL), we highlight the need for stronger theoretical and pedagogical foundations by exploring cognitive-affective mediators (e.g., cognitive load) and learner-related moderators (e.g., learning styles).