We performed a systematic review and meta-analysis of SARS-CoV-2 serology studies to estimate time-dependent sensitivity for a large set of serological assays. We find that seroreversion depends strongly on the antigen and the analytic technique of the assay. We estimate the average time-varying sensitivity for different types of assays, as defined by their technical characteristics, which can be used to approximate the sensitivity of out-of-sample assays. We find large variability between types of assays, with assays of certain characteristics (nucleocapsid antigen, lateral flow assay) having an average of 20% sensitivity at 6 months, while others (receptor-binding domain antigen, sandwich design) have 98% sensitivity. Importantly, we estimate that almost a third of included assays depart significantly from manufacturer specified sensitivity after 6 months. Our analysis outlines the importance of monitoring the seroreversion of serological tests, and provides a tool for researchers to account for this phenomenon.