Background: A Social Story™ (Carol Gray) is a child-friendly intervention that is used to give children with autism spectrum disorders (ASDs) social information in situations where they have social difficulties. Limited evidence mainly using single-case designs suggests that they can reduce anxiety and challenging behaviour. Objectives: The objectives were to conduct a systematic review, use this to develop a manualised intervention and run a feasibility trial to inform a fully powered randomised controlled trial (RCT) on their clinical effectiveness and cost-effectiveness in schools. Design: This is a three-stage study following the Medical Research Council framework for complex interventions. Specifically, it involved a theoretical phase, a qualitative stage and a feasibility trial stage. Setting: Qualitative interviews and focus groups took place in Child and Adolescent Mental Health Service and primary care settings. The feasibility study took place in 37 local mainstream schools. Participants: Fifty children (aged 5-15 years) in mainstream school settings with a diagnosis of ASD were entered into the trial. For each child, an associated teacher and parent was also recruited. Interventions: The intervention was a goal-setting session followed by a manualised toolkit (including a training session) for creating Social Stories™ for use with school-aged children. The comparator treatment was a goal-setting session followed by an attention control. Both arms received treatment as usual. Main outcome measures: Outcomes tested as part of the feasibility study included child-and proxy-completed questionnaires for mental health, quality of life and goal-based outcome measures. Adults additionally completed behaviour diaries and the parental stress index. Results: The review found that the research into social stories is predominantly based in the USA, carried out in under-12-year-olds and using single-case designs. Most studies either did not follow established Social Story criteria or did not report if they did. The assessment of effectiveness presents a largely positive picture but is limited by methodological issues. There were no adequate RCTs and insufficient information to assess a number of important sources of potential bias in most studies. A manualised intervention was produced using an iterative process between user focus groups and a writing team, and assessed in the feasibility study. All 50 participant groups were recruited within the study time frame. Two outcome measures, the Social Responsiveness Scale-2 and the custom-made goal-based measure, showed high levels of completion rates and appeared to be capturing social and behaviour skills targeted by the use of Social Stories. Detailed recommendations for a full trial are provided. Limitations: Blinding of participants was not feasible. Treatment fidelity was not assessed because of low levels of story return rates. Conclusions: The study showed that a fully powered RCT is feasible with an extended geographical footprint. A large amount of data and information has helped to inform the design of this RCT, which will be the subject of a future research grant application. Future work could focus on developing an appropriate blinded outcome measure for this population.