In the assessment of Phanerozoic marine global biodiversity, there has been longstanding interest in quantifying compositional similarities among sampling points as a function of their distances from one another (geodisparity). Previous research has demonstrated that faunal similarity between any two locations tends to decrease significantly as the great circle distance (GCD) between the locations increases, but the rate of decrease begins to stabilize at transoceanic distances. The accuracy of these assessments, and comparisons among different temporal intervals, may suffer, however, because of intervening landmasses that are not accounted for when distance is calibrated simply as GCD. Here, we present a new method for determining the shortest overwater distance (WD) between two marine locations, and we use the method to recalibrate for several Phanerozoic intervals previous measures of global geodisparity in the taxonomic compositions of marine biotas. WD was determined by using a cost-distance approach in ArcGIS, modified to work on a spherical, as opposed to a planar, surface. Results demonstrate two notable effects of using WD. First, mean compositional similarity between locations tends to decrease more continuously as a function of distance with WD than with GCD. Second, pairs of locations with WDs that are at least 50% greater than their GCDs tend to have lower compositional similarity to one another than those with more closely matching WDs and GCDs. These differences are expected as WD better represents the “true” distance between locations; they diminish at GCDs of 5000 km or more when clear, transoceanic paths between locations become more common. Despite these effects, using WD does not alter fundamental temporal trends in global geodisparity through the Phanerozoic observed in previous research, but it is likely to have more significant ramifications for more confined paleobiogeographic investigations.