Variation in observed global generic richness over the Phanerozoic must be partly explained by changes in the numbers of fossils and their geographic spread over time. The influence of sampling intensity (i.e., the number of samples) has been well addressed, but the extent to which the geographic distribution of samples might influence recovered biodiversity is comparatively unknown. To investigate this question, we create models of genus richness through time by resampling the same occurrence dataset of modern global biodiversity using spatially explicit sampling intensities defined by the paleo-coordinates of fossil occurrences from successive time intervals. Our steady-state null model explains about half of observed change in uncorrected fossil diversity and a quarter of variation in sampling-standardized diversity estimates. The inclusion in linear models of two additional explanatory variables associated with the spatial array of fossil data (absolute latitudinal range of occurrences, percentage of occurrences from shallow environments) and a Cenozoic step increases the accuracy of steady-state models, accounting for 67% of variation in sampling-standardized estimates and more than one-third of the variation in first differences. Our results make clear that the spatial distribution of samples is at least as important as numerical sampling intensity in determining the trajectory of recovered fossil biodiversity through time and caution against the overinterpretation of both the variation and the trend that emerge from analyses of global Phanerozoic diversity.