Meta-analyses support the efficacy of cognitive–behavioural therapy (CBT) for schizophrenia in western cultures. This study aimed to compare the efficacy of CBT and supportive therapy (ST) for patients with schizophrenia in China. A multicentre randomized controlled, single-blinded, parallel-group trial enrolled a sample of 192 patients with schizophrenia. All patients were offered 15 sessions of either CBT or ST over 24 weeks and followed up for an additional 60 weeks. All measures used were standardized instruments with good reliability and validity. The Positive and Negative Syndrome Scale (PANSS) was used to assess symptoms of schizophrenia. The Schedule for Assessing Insight (SAI) was used to assess patients’ insight and the Personal and Social Performance Scale (PSP) was used to assess their social functioning. Effect-size analysis showed that patients made rapid improvements in all symptoms, insight and social functioning as measured by the PANSS, SAI and PSP at 12 and 24 weeks and maintained these improvements over the course of the study to 84 weeks. Patients in the CBT group showed significantly greater and more durable improvement in PANSS total score (p = 0.045, between-group d = 0.48), positive symptoms (p = 0.018, between-group d = 0.42) and social functioning (p = 0.037, between-group d = 0.64), with significant differences emerging after completion of therapy. Both CBT and ST combined with medication had benefits on psychopathology, insight and social functioning of patients with schizophrenia. CBT was significantly more effective than ST on overall, positive symptoms and social functioning of patients with schizophrenia in the long term.