Models predicting the distributions of wildlife have become a popular tool in conservation biology and ecology. Their uses are many and varied, including insights into competition theory, predicting the impacts of climate change, and identifying the best locations for protected sites. Predictive modelling requires well distributed data sets and it is no surprise that researchers are increasing turning to remote sensing as a source of predictor variables. Remotely sensed data are well-suited to this application, the full grid of numerical reflectance values or derived indices providing an apparently ideal input to statistical models. The enthusiastic uptake of remotely sensed data in distribution models is not without problems, however, and little attention has been paid to the problems associated with using such data. In this paper, we briefly review the many remote sensing products that are available to distribution modellers and provide examples of their use. We then examine in more detail the assumptions made in using satellite and airborne imagery and the impact the choice of spatial resolution has on the collection of associated field data and the analyses that may be performed. A point of major concern is the mis-registration of data from different sources and we explore the interactions between co-registration, spatial scale and model performance.