Background Evidence for the efficacy of cognitive—behavioural therapy for schizophrenia is promising but evidence for clinical effectiveness is limited. Aims To test the effectiveness of cognitive—behavioural therapy delivered by clinical nurse specialists in routine practice. Method Of 274 referrals, 66 were allocated randomly to 9 months of treatment as usual (TAU), cognitive—behavioural therapy plus TAU (CBT) or supportive psychotherapy plus TAU (SPT) and followed up for 3 months. Results Treatment effects were modest but the CBT condition gave significantly greater improvement in overall symptom severity than the SPT or TAU conditions combined (F (1,53)=4.14; P=0.05). Both the CBT and SPT conditions combined gave significantly greater improvement in severity of delusions than did the TAU condition (F (1,53)=4.83; P=0.03). Clinically significant improvements were achieved by 7/21 in the CBT condition (33%), 3/19 in the SPT condition (16%) and 2/17 in the TAU condition (12%). Conclusions Cognitive—behavioural therapy delivered by clinical nurse specialists is a helpful adjunct to routine care for some people with chronic psychosis.