Cyclists are frequent casualties in road traffic collisions, and cyclist inattention is often implicated. Children are more distractible than adults and so are arguably more vulnerable road users, including when cycling. Research shows that misallocation of attention or distraction is a common factor in crashes involving cyclists, and child cyclists are more likely to be seriously injured, possibly because their hazard perception is inferior to that of adults. Video-based training paradigms designed to improve children’s hazard perception and situation awareness have yielded mixed findings, but none have examined positive transfer of such training interventions to on-road cycling performance. The aim of this study was to use an immersive video-based training protocol to improve attention allocation, situation awareness and on-road cycling performance in children who had completed UK cycle training aimed at 9–11-year-olds. Thirty-three children aged 10–12 years were randomly allocated to either an Intervention group or a Control group. All participants reported their cycling behaviour and cycling self-efficacy, and completed online situation awareness tests at baseline, post-test, and retention stages. Between baseline and post-test, the intervention group (n = 17) completed a lab-based training protocol in which they viewed real-world cyclist point-of-view footage on an immersive screen while pedalling on a stationary cycle. As they navigated five virtual routes their task was to demonstrate awareness of potential hazards and other information pertinent to their safety, with decreasing levels of support from the researcher as they progressed. The control group did not receive the training intervention. All participants’ cycling performance was individually assessed on urban roads on two testing occasions, by qualified cycle instructors. The intervention group outperformed the control group in terms of their on-road performance, in terms of Making good and frequent observations (Observation; F[1,31]) = 16.53, ηp2 = 0.35, p = 0.0003,); Communicating intentions clearly to others (Communication; F[1,31]) = 13.70, ηp2 = 0.31, p = 0.001); Choosing and maintaining the most suitable riding positions (Position; F[1,31]) = 7.41, ηp2 = 0.19, p = 0.01); and Understanding priorities on the road, particularly at junctions (Priorities; F[1,31]) = 6.88, ηp2 = 0.18, p = 0.01), although this was not accompanied by significant changes in their cycling self-efficacy and situation awareness, all p’s > 0.05. The present findings suggest that an immersive video-based protocol such as the one herein may improve children’s safety when cycling on roads and may therefore be an effective complement to current cycle training protocols. Given ongoing governmental investment in cycling infrastructure in the UK, with a concomitant increase in micromobility modes of active travel, national policy will need to consider the additional perceptual-cognitive demands that will face young cyclists; immersive training is a potentially efficient and cost-effective way to mitigate these demands. However, further research attention should be devoted to the development of effective situation awareness tests – preferably ones that correlate with, if not predict, children’s independent on-road cycling performance.