We conducted a meta-analysis to investigate the HPA axis response to social stress in studies that used the Trier Social Stress Test (TSST), or comparable distressing paradigms, in individuals with either depression or schizophrenia. Sample size-adjusted effect sizes (Hedge's g statistic) were calculated to estimate the HPA axis stress response to social stress. We used a meta-regression model to take into account the moderating effect of the baseline cortisol level. Participants with depression show an activation pattern to social stress similar to that of healthy controls. Despite a normal cortisol production rate, individuals with schizophrenia have lower cortisol levels than controls both in anticipation and after exposure to social stress. Participants with depression and higher cortisol levels before the task have an increased cortisol production and reached higher cortisol levels during the task. This may be explained by the presence of an impaired negative feedback. The activation pattern present in schizophrenia may explain the reduced ability to appropriately contextualize past experiences shown by individuals with psychosis in social stressful situation.