Background: Cognitive behavioural therapy (CBT) is a recommended treatment for people with psychosis, with meta-analyses showing important benefits when compared to treatment as usual (TAU). However, there has been growing debate as to whether CBT has specific benefits over and above those attributable to less intensive psychosocial interventions, such as befriending and supportive counselling. Recent meta-analyses examining this question have suffered from various difficulties including potential bias and error. After detailing these problems, we present a protocol for a new review. Methods and design: The Cochrane Central Register of Controlled Trials (CENTRAL), PubMed, EMBASE, previous reviews and online trial registers will be systematically searched for randomised trials examining the efficacy and safety of CBT for psychosis, when compared to (a) TAU and (b) other psychosocial treatments. Primary outcomes are symptomatic change and numbers achieving clinically significant improvement. Secondary outcomes include adverse effects, change in target complaint, subjectively defined recovery and relapse. Discussion: This review will provide service users, relatives and policy-makers with a comprehensive and up-to-date assessment of the efficacy and safety of CBT for psychosis. Advance publication and peer-review of hypotheses and methods should maximise transparency, enhancing the reliability and quality of the findings.