Understanding the neural mechanisms underlying psychological therapy could aid understanding of recovery processes and help target treatments. The dual-process model hypothesises that psychological therapy is associated with increased emotional-regulation in prefrontal brain regions and decreased implicit emotional-reactivity in limbic regions; however, research has yielded inconsistent findings. Meta-analyses of brain activity changes accompanying psychological therapy (22 studies, n = 352) and neural predictors of symptomatic improvement (11 studies, n = 293) in depression and anxiety were conducted using seed-based d mapping. Both resting-state and task-based studies were included, and analysed together and separately. The most robust findings were significant decreases in anterior cingulate/paracingulate gyrus, inferior frontal gyrus and insula activation after therapy. Cuneus activation was predictive of subsequent symptom change. The results are in agreement with neural models of improved emotional-reactivity following therapy as evidenced by decreased activity within the anterior cingulate and insula. We propose compensatory as well as corrective neural mechanisms of action underlie therapeutic efficacy, and suggest the dual-process model may be too simplistic to account fully for treatment mechanisms. More research on predictors of psychotherapeutic response is required to provide reliable predictors of response.