High-resolution gridded precipitation datasets provide a valuable, spatially-complete picture of historical precipitation. However, gridded data are often treated as “truth” without acknowledging the significant assumptions and simplifications need to turn point observations into a spatially-complete gridded product. Here extreme precipitation is evaluated in five frequently-used gridded precipitation datasets by comparing with Global Historical Climatology Network station data over California. Substantial differences are found among the datasets and can be traced back to differing assumptions about how precipitation varies in space.