Various clinical guidelines recommend cognitive behavioural therapy (CBT) to treat psychosis without reference to patients' thought disorder. However, there is a risk that disorganized thinking hampers CBT. We tested the prediction that thought disorder would interfere with the effectiveness of CBT for hallucinations and delusions, compared to treatment as usual and supportive counselling, in secondary data from two large, single blind randomised controlled trials. We fitted latent growth curve models separately for the development of frequency and distress of symptoms. CBT was significantly more successful than counselling in reducing delusional frequency in the short term and hallucinatory distress at any point, even in those with relatively high thought disorder. We found little evidence that clinicians should restrict CBT in this subgroup of patients. Nevertheless, the findings highlight the importance of effective initial treatment of thought disorder in maximising the benefit of CBT for psychosis, particularly for reducing distress from hallucinations.