Abstract Depression has an annual prevalence of 1–6% in the community; 50–60% of depressed individuals might not respond to conventional pharmacotherapy. Transcranial magnetic stimulation (TMS) non-invasively stimulates superficial cortex in patients, for investigative and therapeutic purposes. It is usually applied over the prefrontal cortex at frequencies of 1–20 Hz at motor threshold intensity. We present a meta-analysis of 24 studies evaluating the antidepressant effect of TMS for major depressive or bipolar disorder in treatment groups ≥10 patients. Out of 617 patients receiving active rTMS, 218 (35.3%) were classified as ‘responders’, whereas only 71 (13.1%) of 543 patients undergoing sham rTMS met the criteria for clinical response. The Peto odds ratio meta-analysis indicated that this difference is statistically significant, with an odds ratio of 3.88 (95%-CI: 2.94–5.13). Heterogeneity between studies did not exceed that expected by chance and there was no significant publication bias. Based on these data, five patients (95% CI = 4–6) need to be treated in order to obtain a clinical response attributable to rTMS, a respectable effect size among psychiatric (add-on) treatments. Unfortunately, there is no compelling evidence regarding the most effective combination of rTMS parameters. The literature indicates that future trials should employ a greater number of rTMS sessions, adequate concealment allocation and an individualized approach to locating the DLPFC using neuroimaging. Also, more knowledge is needed regarding the characteristics of patients who benefit from this treatment and the size and persistence of clinical effects.