Wetland restoration is a popular nutrient management strategy for improving water quality in agricultural catchments. However, a wetland’s ability to retain phosphorus is highly variable and wetlands can sometimes be a source of phosphorus to downstream ecosystems. Here, we used a meta-analysis approach to explore the source and sink capacity of 139 wetlands for both total phosphorus (TP) and the more bioavailable form, phosphate (PO4 3−), at seasonal and annual timescales. Median retention efficiency across all studies is 32% for TP and 28% for PO4 3−, however the range is extremely broad. We found that wetlands are often sinks for TP (84% of site-years) and PO4 3− (75% of site years). The median TP retention within wetlands that are sinks (2.0 g·m−2 ·yr−1 ) is greater than release by wetlands that are sources (−0.5 g·m−2 ·yr−1 ). In contrast, for PO4 3−, median retention within wetlands that are phosphorus sinks (0.8 g·m−2 ·yr−1 ) is of similar magnitude to that released by wetlands that are phosphorus sources (−0.7 g·m−2 ·yr−1 ). We found that phosphorus release from wetlands coincides with higher hydraulic loading rates, lower influent phosphorus concentration, and legacy soil/sediment phosphorus. Phosphate releases were especially common in wetlands used for treating municipal wastewater, as well as restored and constructed wetlands with flashy, precipitation-driven flow. We found that experimental design may inherently bias our understanding of wetland performance for phosphorus retention as studies conducted in mesocosms outperform other wetland types. Analysis of monthly data demonstrated significant temporal variability in wetland phosphorus dynamics, often switching from retention to release many times within a year, but with no generalizable seasonal trends. Our results highlight the value of restoring wetlands for phosphorus retention and point to ways of furthering their utility towards improving water quality by simultaneously targeting retention enhancing measures and release avoidance.