Background <p>Limitations of traditional standardized patients exist with regard to psychiatric interview training, and there is a pressing need to develop more effective tools. The aim of this study is to develop and conduct a preliminary evaluation of a virtual standardized patient (VSP) system designed for this purpose.</p> Methods <p>This study comprised two primary components: (1) the development of a VSP system for psychiatry interviews; and (2) a randomized controlled trial (RCT) to evaluate its effectiveness. An experienced psychiatrist drafted the initial case interview scripts that were subsequently refined using a modified Delphi method to establish the final versions. The VSP system was then constructed using a large language model (LLM) of ERNIE 3.5–8&#xa0;K. We then recruited 20 valid medical students who were randomly allocated to either the intervention group or control group for the RCT. Both groups completed a 20 class hours training. A participant’s clinical interview skills were assessed before and after the intervention, while the intervention group additionally provided feedback on their VSP user experience.</p> Results <p>The developed VSP system incorporated three psychiatric cases. It demonstrated good consistency upon expert evaluation. After the intervention, the VSP group demonstrated significantly higher clinical interview scores compared to the controls. Additionally, the participants rated the VSP system as having good usability, with particular strengths in operational ease and effectiveness for clinical interview training.</p> Conclusions <p>In this study, we developed a VSP system for psychiatric interview training. The preliminary validation demonstrated its efficacy for enhancing medical student clinical interview skills. This system provides valuable insights for the development of innovative psychiatric education tools.</p> Clinical trial number <p>Not applicable.</p>