PURPOSE: The study of control beliefs in psychotherapy research has been neglected in the past years. Based on the evidence that some patients do not benefit enough from therapy because of inadequate expectancies regarding the responsibility and the mechanisms of therapeutic change, assessing control beliefs specific to the psychotherapy context and linking them to therapy outcome can help highlighting this specific aspect and reactivating a neglected field of clinical research. METHOD: Using a new validated instrument (Questionnaire on Control Expectancies in Psychotherapy, TBK), this study investigated whether and how perceived responsibility for change predicts favourable response to group cognitive-behavioural therapy in a sample of 49 outpatients with social anxiety disorder (SAD). Patient engagement and therapy-related self-efficacy were assessed as possible process variables. RESULTS: Among therapy-related control beliefs, low powerful others expectancies (towards the therapist) were found to be the strongest predictor for clinical improvement at follow-up. At a process level, analyses of mediation showed that powerful others expectancies predicted therapy engagement, which then influenced the degree of clinical improvement on social anxiety levels and global symptoms. The association between therapy-specific internality and outcome was confirmed for social anxiety at follow-up and was partially mediated by therapy-related self-efficacy. CONCLUSIONS: Findings confirm that therapy-related control beliefs predict psychotherapy process (patient engagement and therapy-specific self-efficacy) and outcome in cognitive-behavioural group therapy for SAD. Implications for clinicians and for future research are discussed.