Background Although music therapy (MT) is considered an effective intervention for young people with mental health needs, its efficacy in clinical settings is unclear. We therefore examined the efficacy of MT in clinical practice. Methods Two hundred and fifty‐one child (8–16 years, with social, emotional, behavioural and developmental difficulties) and parent dyads from six Child and Adolescent Mental Health Service community care facilities in Northern Ireland were randomised to 12 weekly sessions of MT plus usual care [n = 123; 76 in final analyses] or usual care alone [n = 128; 105 in final analyses]. Follow‐up occurred at 13 weeks and 26 weeks postrandomisation. Primary outcome was improvement in communication (Social Skills Improvement System Rating Scales) (SSIS) at 13 weeks. Secondary outcomes included social functioning, self‐esteem, depression and family functioning. Results There was no significant difference for the child SSIS at week 13 (adjusted difference in mean 2.4; 95% CI −1.2 to 6.1; p = .19) or for the guardian SSIS (0.5; 95% CI −2.9 to 3.8; p = .78). However, for participants aged 13 and over in the intervention group, the child SSIS communication was significantly improved (6.1, 95% CI 1.6 to 10.5; p = .007) but not the guardian SSIS (1.1; 95% CI −2.9 to 5.2; p = .59). Overall, self‐esteem was significantly improved and depression scores were significantly lower at week 13. There was no significant difference in family or social functioning at week 13. Conclusions While the findings provide some evidence for the integration of music therapy into clinical practice, differences relating to subgroups and secondary outcomes indicate the need for further study. ISRCTN Register; ISRCTN96352204.