Clients and therapists often have different perspectives on their therapeutic alliance (TA), affecting the process and outcome of therapy. The aim of the present meta-analysis was to assess the mean differences between clients’ and therapists’ estimations of TA among clients with severe disturbances, while focusing on two potential moderators: client diagnosis and alliance instrument.

We conducted a systematic literature search of studies examining both client perspective and therapist perspective on TA in psychotherapy among people with schizophrenia spectrum disorders, personality disorders, and substance misuse disorders. We then analyzed the data using a random-effects meta-analytic model with Cohen’s d standardized mean effect size.

Heterogeneity analyses (k = 22, Cohen’s d = −.46, 95% confidence interval = .31–1.1) produced a significant Q-statistic (Q = 94.96) and indicated high heterogeneity, suggesting that moderator analyses were appropriate.

Our findings show that the type of TA instrument moderates the agreement on TA between client and therapist, but there was no indication of the client’s diagnosis moderating the effect. The agreement between client and therapist estimations seems to be dependent on the instrument that is used to assess TA. Specific setting-related instruments seem to result in higher agreement between clients’ and therapists’ estimations than do more general instruments that are applied to assess TA.