BACKGROUND: Little evidence is available for head-to-head comparisons of psychosocial interventions and pharmacological interventions in psychosis. We aimed to establish whether a randomised controlled trial of cognitive behavioural therapy (CBT) versus antipsychotic drugs versus a combination of both would be feasible in people with psychosis. METHODS: We did a single-site, single-blind pilot randomised controlled trial in people with psychosis who used services in National Health Service trusts across Greater Manchester, UK. Eligible participants were aged 16 years or older; met ICD-10 criteria for schizophrenia, schizoaffective disorder, or delusional disorder, or met the entry criteria for an early intervention for psychosis service; were in contact with mental health services, under the care of a consultant psychiatrist; scored at least 4 on delusions or hallucinations items, or at least 5 on suspiciousness, persecution, or grandiosity items on the Positive and Negative Syndrome Scale (PANSS); had capacity to consent; and were help-seeking. Participants were assigned (1:1:1) to antipsychotics, CBT, or antipsychotics plus CBT. Randomisation was done via a secure web-based randomisation system (Sealed Envelope), with randomised permuted blocks of 4 and 6, stratified by gender and first episode status. CBT incorporated up to 26 sessions over 6 months plus up to four booster sessions. Choice and dose of antipsychotic were at the discretion of the treating consultant. Participants were followed up for 1 year. The primary outcome was feasibility (ie, data about recruitment, retention, and acceptability), and the primary efficacy outcome was the PANSS total score (assessed at baseline, 6, 12, 24, and 52 weeks). Non-neurological side-effects were assessed systemically with the Antipsychotic Non-neurological Side Effects Rating Scale. Primary analyses were done by intention to treat; safety analyses were done on an as-treated basis. The study was prospectively registered with ISRCTN, number ISRCTN06022197. FINDINGS: Of 138 patients referred to the study, 75 were recruited and randomly assigned-26 to CBT, 24 to antipsychotics, and 25 to antipsychotics plus CBT. Attrition was low, and retention high, with only four withdrawals across all groups. 40 (78%) of 51 participants allocated to CBT attended six or more sessions. Of the 49 participants randomised to antipsychotics, 11 (22%) were not prescribed a regular antipsychotic. Median duration of total antipsychotic treatment was 44·5 weeks (IQR 26-51). PANSS total score was significantly reduced in the combined intervention group compared with the CBT group (-5·65 [95% CI -10·37 to -0·93]; p=0·019). PANSS total scores did not differ significantly between the combined group and the antipsychotics group (-4·52 [95% CI -9·30 to 0·26]; p=0·064) or between the antipsychotics and CBT groups (-1·13 [95% CI -5·81 to 3·55]; p=0·637). Significantly fewer side-effects, as measured with the Antipsychotic Non-neurological Side Effects Rating Scale, were noted in the CBT group than in the antipsychotics (3·22 [95% CI 0·58 to 5·87]; p=0·017) or antipsychotics plus CBT (3·99 [95% CI 1·36 to 6·64]; p=0·003) groups. Only one serious adverse event was thought to be related to the trial (an overdose of three paracetamol tablets in the CBT group). INTERPRETATION: A head-to-head clinical trial of CBT versus antipsychotics versus the combination of the two is feasible and safe in people with first-episode psychosis. FUNDING: National Institute for Health Research.