The clinical benefit of convalescent plasma (CP) for patients with coronavirus disease (COVID)-19 is still debated. In this systematic review and meta-analysis, we selected 10 randomized clinical trials (RCTs) and 15 non-randomized studies (total number of patients = 22,591) of CP treatment and evaluated two different scenarios: (1) disease stage of plasma recipients and (2) donated plasma antibody titer, considering all-cause mortality at the latest follow-up. Our results show that, when provided at early stages of the disease, CP significantly reduced mortality: risk ratio (RR) 0.72 (0.68, 0.77), p < 0.00001, while provided in severe or critical conditions, it did not (RR: 0.94 [0.86, 1.04], p = 0.22). On the other hand, the benefit on mortality was not increased by using plasma with a high-antibody titer compared with unselected plasma. This meta-analysis might promote CP usage in patients with early-stage COVID-19 in further RCTs to maximize its benefit in decreasing mortality, especially in less affluent countries.