OBJECTIVES:The efficacy of cognitive behaviour therapies for psychosis (CBTp) has been sufficiently established for its inclusion in some national treatment guidelines. However, treatment efficacy does not guarantee effectiveness in routine practice, where clinician expertise and patient mix may be different. Thus, we evaluated the applicability, acceptability and effectiveness of CBTp when offered routinely in a public mental health service. DESIGN:A prospectively recruited representative sample (N=94) of patients with psychotic disorders from a geographic catchment area in Melbourne, Australia, was randomized to CBTp or treatment as usual. METHOD:The CBTp intervention included psychoeducation, positive symptom, and co-morbid problem components. Therapists collaboratively negotiated goals with patients and utilized treatment components from a manual. Intention-to-treat analyses used data at baseline, 9 months and 18 months. RESULTS:Working alliance and client satisfaction measures indicated excellent acceptability. Both groups improved to a similar degree on the Positive And Negative Syndrome Scale, with no advantage for the CBTp group. The most frequent CBTp components implemented were co-morbid disorders/personal issues and personalized psychoeducation; working with persisting symptoms was the main focus in relatively few sessions. CONCLUSIONS:The lack of advantage for the CBTp group may be due to: the infrequent focus on positive symptoms (the most replicated outcome domain for CBTp); the unselected nature of the patients (rather than those with distressing symptoms or referred by clinicians); or, to insufficient therapist expertise.