Ecological inferences are often based on the locations at which species are present, but many species records have substantial uncertainty in spatial metadata, limiting their utility for fine-scale analyses. This is especially prevalent in historical records such as museum specimens, and in some citizen-science data. For example, the North American Breeding Bird Survey (BBS) has 55+ years of bird data from regular transects (“routes”) across the continent but was not designed to capture the spatial component of point count events, limiting analyses of species-habitat relationships for which it would otherwise be well suited. We present a new methodology for quantifying location uncertainty in BBS records using digitized estimated stop locations, deriving the corresponding environmental covariate uncertainty distributions, and incorporating this information into hierarchical species distribution models using informative Bayesian priors. This approach allows for estimation of species–environment relationships in a way that fully accounts for underlying spatial uncertainty. We quantify stop-location uncertainty in BBS data across the central United States, model bird–land cover relationships in the upper Midwest, and validate our method by comparing posterior land cover estimates to known covariate values for a subset of GPS-digitized stop locations. We provide code for implementing this method in R. Posterior land cover estimates (forest, grass/hay, and developed land cover), based on our informative priors, were highly correlated with known land cover values from GPS-digitized stop locations. Our approach thus makes it possible to responsibly leverage large historic and citizen science databases, such as the BBS, for fine-scale ecological analyses.