Young people with social disability and non-psychotic severe and complex mental health problems are an important group. Without intervention, their social problems can persist and have large economic and personal costs. Thus, more effective evidence-based interventions are needed. Social recovery therapy is an individual therapy incorporating cognitive-behavioural techniques to increase structured activity as guided by the participant's goals. This trial aimed to test whether or not social recovery therapy provided as an adjunct to enhanced standard care over 9 months is superior to enhanced standard care alone. Enhanced standard care aimed to provide an optimal combination of existing evidence-based interventions. A pragmatic, single-blind, superiority randomised controlled trial was conducted in three UK centres: Sussex, Manchester and East Anglia. Participants were aged 16-25 years with persistent social disability, defined as < 30 hours per week of structured activity with social impairment for at least 6 months. Additionally, participants had severe and complex mental health problems, defined as at-risk mental states for psychosis or non-psychotic severe and complex mental health problems indicated by a Global Assessment of Functioning score ≤ 50 persisting for ≥ 6 months. Two hundred and seventy participants were randomised 1 : 1 to either enhanced standard care plus social recovery therapy or enhanced standard care alone. The primary outcome was weekly hours spent in structured activity at 15 months post randomisation. Secondary outcomes included subthreshold psychotic, negative and mood symptoms. Outcomes were collected at 9 and 15 months post randomisation, with maintenance assessed at 24 months. The addition of social recovery therapy did not significantly increase weekly hours in structured activity at 15 months (primary outcome treatment effect -4.44, 95% confidence interval -10.19 to 1.31). We found no evidence of significant differences between conditions in secondary outcomes at 15 months: Social Anxiety Interaction Scale treatment effect -0.45, 95% confidence interval -4.84 to 3.95; Beck Depression Inventory-II treatment effect -0.32, 95% confidence interval -4.06 to 3.42; Comprehensive Assessment of At-Risk Mental States symptom severity 0.29, 95% confidence interval -4.35 to 4.94; or distress treatment effect 4.09, 95% confidence interval -3.52 to 11.70. Greater Comprehensive Assessment of At-Risk Mental States for psychosis scores reflect greater symptom severity. We found no evidence of significant differences at 9 or 24 months. Social recovery therapy was not estimated to be cost-effective. The key limitation was that missingness of data was consistently greater in the enhanced standard care-alone arm (9% primary outcome and 15% secondary outcome missingness of data) than in the social recovery therapy plus enhanced standard care arm (4% primary outcome and 9% secondary outcome missingness of data) at 15 months. We found no evidence for the clinical superiority or cost-effectiveness of social recovery therapy as an adjunct to enhanced standard care. Both arms made large improvements in primary and secondary outcomes. Enhanced standard care included a comprehensive combination of evidence-based pharmacological, psychotherapeutic and psychosocial interventions. Some results favoured enhanced standard care but the majority were not statistically significant. Future work should identify factors associated with the optimal delivery of the combinations of interventions that underpin better outcomes in this often-neglected clinical group. Current Controlled Trials ISRCTN47998710. This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment Vol. 25, No. 70. See the NIHR Journals Library website for further project information.