Imagery training with adult athletes is widely used to improve performance. One underlying mechanism is the optimization of mental movement representations. However, past research has focused mainly on adults and has left open for further research on whether imagery also improves mental representations and performance in young athletes. The present study examined these questions in a sample of 56 female gymnasts aged 7 to 15 years. In a cross-over experimental design (imagery first vs. imagery last), regular training with imagery was compared with regular training only in high- versus low-expertise athletes. The 4-week long imagery training had positive effects on performance only for the high-expertise athletes in the imagery-last condition. The results of the Structural Dimensional Analysis of Mental Representation method regarding changes in the mental representations were inconsistent. Thus, imagery training can promote motor learning in young athletes only under some conditions. We discuss possible reasons for the heterogeneous results and ways for improving the strength and reliability of the intervention effects.