GPS sensors have an inherent positional uncertainty that is often neglected in environmental modeling. In this article we study the propagation of positional uncertainty in grid-based geo-information systems. The probability is obtained that a point has an actual position outside the raster cell in which it was observed. For multiple points, these probabilities serve as weights to update a raster value and adjust for positional uncertainty. We show the effect of positional uncertainty propagation by means of simulations using rasters with different levels of spatial autocorrelation, as well as an illustration of a real-world example in which we estimate exposure to air pollution. We found that the propagated uncertainty was highest when the spatial autocorrelation is low, with a root mean squared error of 7% compared to 1% for the high spatial autocorrelation scenario. We conclude that positional uncertainty propagates through environmental models and propose a simple probabilistic method to account for it.