Background Cognitive remediation (CR) preceding cognitive-behavioural therapy for psychosis (CBTp) was trialled within routine clinical services, with the hypothesis that following first-episode non-affective psychosis CR would enhance CBTp efficacy by improving neuropsychological performance. Method A total of 61 patients with DSM-IV non-affective psychoses waiting for routine CBTp were randomized to computerized CR over 12 weeks, supported by a trained support worker, or time-matched social contact (SC). Primary outcome was the blind-rated Psychotic Symptoms Rating Scale (PSYRATS). Secondary outcomes included measures of CBTp progress, cognition, symptoms, insight and self-esteem: all at baseline, after CR (12 weeks) and after CBTp (42 weeks). PSYRATS and global neuropsychological efficacy were tested using mixed-effects models with a groupÂ ×Â time interaction term. Measures of CBTp progress and some neuropsychological measures were modelled by regression. Results There was no significant difference between the CR and SC groups in PSYRATS (groupÂ ×Â time coefficient 0.3, 95% confidence interval-0.4 to 1.1, pÂ =Â 0.39). However, after CR CBTp was shorter [median 7 sessions, interquartile range (IQR) 2-12 after CR; median 13, IQR 4-18 after SC; model pÂ =Â 0.011] and linked to better insight (pÂ =Â 0.02). Global cognition did not improve significantly more after CR (pÂ =Â 0.20) but executive function did (Wisconsin Card Sort, pÂ =Â 0.012). Conclusions CBTp courses preceded by CR were far shorter but achieved the same outcome as CBTp preceded by an active control, consistent with neuropsychological improvement enhancing CBTp. CR was delivered by staff with minimal training, offering the potential to reduce the costs of CBTp considerably. © 2013 Cambridge University Press.