Abstract Background High-fidelity patient simulators in team training are becoming popular, though research showing benefits of the training process compared to low-fidelity models is rare. We explored in situ training for paediatric teams in an emergency department using a low-fidelity model (plastic doll) and a high-fidelity paediatric simulator, keeping other contextual factors constant. The goal was to study differences in trainees’ and trainers’ performance along with their individual experiences, during in situ training, using either a low-fidelity model or a high-fidelity paediatric simulator. Methods During a two-year period, teams involved in paediatric emergency care were trained in groups of five to nine. Each team performed one video-recorded paediatric emergency scenario. A case control study was undertaken in which 34 teams used either a low-fidelity model (n = 17) or a high-fidelity paediatric simulator (n = 17). The teams’ clinical performances during the scenarios were measured as the time elapsed to prescribe as well as deliver oxygen. The trainers were monitored regarding frequency of their interventions. We also registered trainees’ and trainers’ mental strain and flow experience. Results Of 225 trainees’ occasions during 34 sessions, 34 trainer questionnaires, 163 trainee questionnaires, and 28 videos, could be analyzed. Time to deliver oxygen was significantly longer (p = 0.014) when a high-fidelity simulator was used. The trainees’ mental strain and flow did not differ between the two types of training. The frequency of trainers interventions was lower (p