Coronavirus disease 2019 (COVID-19) is an infectious disease that can develop multiple complications and even be life-threatening. The aim of this study is to summarize current evidence of C-reactive protein’s (CRP) predictive value for disease severity and survival of COVID-19 patients, focusing on curing patients and reducing the risk of death. We systematically searched related studies from four large databases: Web of Science, PubMed, China National Knowledge Infrastructure (CNKI), and Wanfang Database, all published between December 2019 and June 2021. Then, we implemented meta-analysis using random-effects models through STATA 15.1 and Review Manager 5.3. We also implemented sensitivity analysis and used funnel plots to check publication bias. From the systematic search of the four databases, we were able to identify 18 studies containing a total of 3052 patients. Meta-analysis results showed that 1) CRP levels were lower in non-severe patients than in severe patients (Standardized Mean Difference ( SMD) =  − 0.87 mg/L, 95% Confidence Interval (CI) = [ − 1.27, − 0.47], p < 0.001); 2) CRP levels were lower in non-intensive care unit (ICU) patients than in ICU patients ( SMD = − 1.39 mg/L, 95% CI = [− 1.68, − 1.11], p < 0.001), and 3) CRP levels were lower in survivors than in non-survivors ( SMD =− 1.32 mg/L, 95% CI = [− 1.95, − 0.69], p < 0.001). Sensitivity analysis showed these results were stable. Funnel plots indicated no publication bias. The CRP level may timely reflect disease severity and predict survival of COVID-19 patients and may be worthy of further popularization and application in clinic practice.