News feed function becomes very popular in many social networking services and news aggregators, as it delivers the messages from users' subscribed sources. More recently, location has been introduced to the news feed function, which returns the news items relevant to the user's location. However, with the large number of the news items generated by the sources, existing news feed systems opt to return the top-k most recent ones, which completely overlooks the messages' spatial relevance and may end up in missing more geographically close ones. In this paper, we present GeoRank, an efficient location-aware news feed ranking system that provides top-k new feeds based on (a) spatial proximity, (b) temporal proximity, and (c) user preferences. GeoRank encapsulates spatio-temporal pruning techniques to improve its response time and efficiency. GeoRank is composed of two main modules, namely, query processor and message updater. The query processor module is triggered by the user, upon logging on to the system, to provide the top-k ranked location-based news feeds. The message updater module is a process running in the background, which keeps maintaining statistics used by the query processor module. Extensive experimental results, based on real and synthetic data sets, show the scalability and efficiency of GeoRank.