The potential of virtual reality (VR) as a technology for low-risk and cost-effective professional training in the nuclear sector is increasingly being discovered, for instance regarding the decommissioning of nuclear facilities. However, previous studies and initial systematic literature analyses1,2 have mainly focused on areas of application and, to some extent, the effectiveness of these kinds of trainings but have only paid little attention to their design. According to current theories of multimedia learning3, design factors are particularly important for the effectiveness of such trainings. As part of an ongoing research and development project4, that, inter alia, aims to provide VR-based trainings for nuclear interim storage and nuclear power plant decommissioning, we sought to derive sound principles for their didactic design in the early development stage. For this purpose, we built on findings from an exploratory literature review on fundamental design principles for virtual reality trainings5 and conducted six user trials with nuclear energy professionals using a first VR prototype for the training of safety measures. We used the shadowing method to first observe the users while completing a 30-minute learning scenario in terms of their experience and usage behavior and subsequently conduct in-depth interviews to derive factors for suitable design of such trainings. Based on the results, we created an early model that contains nine dimensions of recommendations for the design of VR trainings to promote learning for instructional designers and immersive media designers, that, for instance, advise them to enable intuitive control and locomotion while considering individual user preferences, provide a sufficient but not distracting degree of realism, reduce VR sickness and negative feelings, and guide immersive learning through supporting spatial orientation and providing precise instructions and feedback. Considering these results, we notice that although they provide important implications for the practical design of VR applications, they barely reveal or address any specifics of the nuclear power sector. For further framework development in this respect, we plan to conduct studies particularly focusing on rather specific design principles of VR trainings for the nuclear sector, that, besides user tests with process-experienced personnel, will further include expert interviews with software developers and users to provide deeper insights.