Ecologists and paleontologists alike are increasingly using the fossil record as a spatial data set, in particular to study the dynamics and distribution of geographic range sizes among fossil taxa. However, no attempts have been made to establish how accurately range sizes and range-size dynamics can be preserved. Two fundamental questions are: Can common paleo range-size reconstruction methods accurately reproduce known species’ ranges from locality (i.e., point) data? And, are some reconstruction methods more reliable than others? Here, we develop a methodological framework for testing the accuracy of commonly used paleo range-size reconstruction methods (maximum latitudinal range, maximum great-circle distance, convex hull, and alpha convex hull) in different extinction-related biogeographic scenarios. We use the current distribution of surface water bodies as a proxy for “preservable area,” in which to test the performance of the four methods. We find that maximum great-circle distance and convex-hull methods most reliably capture changes in range size at low numbers of fossil sites, whereas convex hull performs best at predicting the distribution of “victims” and “survivors” in hypothetical extinction scenarios. Our results suggest that macroevolutionary and macroecological patterns in the relatively recent past can be studied reliably using only a few fossil occurrence sites. The accuracy of range-size reconstruction undoubtedly changes through time with the distribution and area of fossiliferous sediments; however, our approach provides the opportunity to systematically calibrate the quality of the spatial fossil record in specific environments and time intervals, and to delineate the conditions under which paleobiologists can reconstruct paleobiogeographical, macroecological, and macroevolutionary patterns over critical intervals in Earth history.