Word clouds are a popular method to visualize the frequency of words in textual data. Nowadays many text-based data sets, such as Flickr tags, are geo-referenced, that is, they have an important spatial component. However, existing automated methods to generate word clouds are unable to incorporate such spatial information. We introduce geo word clouds: word clouds which capture not only the frequency but also the spatial relevance of words. Our input is a set of locations from one (or more) geographic regions with (possibly several) text labels per location. We aggregate word frequencies according to point clusters and employ a greedy strategy to place appropriately sized labels without overlap as close as possible to their corresponding locations. While doing so we "draw" the spatial shapes of the geographic regions with the corresponding labels. We experimentally explore trade-offs concerning the location of labels, their relative sizes and the number of spatial clusters. The resulting word clouds are visually pleasing and have a low error in terms of relative scaling and locational accuracy of words, while using a small number of clusters per label.