Background Structured violence risk assessment is an essential part of treatment planning for violent young people. The Structured Assessment of Violence Risk in Youth (SAVRY) has been shown to have good reliability and validity in a range of settings but has hardly been studied in adolescent mental health services. Aims This study aimed to evaluate the long‐term predictive validity of the SAVRY in adolescent psychiatry settings. Methods In a prospective study, 200 SAVRY assessments of adolescents were acquired from psychiatric, forensic and correctional settings. Re‐offending records from the Finnish National Crime Register were collected. Receiver operating curve statistics were applied. Results High SAVRY total and individual subscale scores and low values on the protective factor subscale were significantly associated with subsequent adverse outcomes, but the predictive value of the total score was weak. At the risk item level, those indicating antisocial lifestyle, absence of social support and pro‐social involvement were strong indicators of subsequent criminal convictions, with or without violence. The SAVRY summary risk rating was the best indicator of likelihood of being convicted of a violent crime. After allowing for sex, age, psychiatric diagnosis and treatment setting, for example, conviction for a violent crime was over nine times more likely among those young people given high SAVRY summary risk ratings. Conclusions The SAVRY is a valid and useful method for assessing both short‐term and long‐term risks of violent and non‐violent crime by young people in psychiatric as well as criminal justice settings, adding to a traditional risk‐centred assessment approach by also indicating where future preventive treatment efforts should be targeted. The next steps should be to evaluate its role in everyday clinical practice when using the knowledge generated to inform and monitor management and treatment strategies. Copyright © 2014 John Wiley & Sons, Ltd.