Social media represents an emerging source of big data with rich mobility information embedded. While extensive studies in cartography, geographic information science, and visualization have been conducted to extract movement patterns from spatial data, new challenges and opportunities arise for deriving geographic patterns of mobility at scale from massive social media data. Conventional methods (e.g. flow mapping) cannot effectively capture geographic features in the process of addressing visualization concerns such as visual clutter. Furthermore, these methods are not scalable to the volume and velocity of social media data. As a consequence, geographic attributes of mobility (e.g. for representing movement trajectories of social media users) are not adequately captured for better understanding actual mobility patterns (e.g. in disease transmission and road traffic). To address this problem, this paper describes a scalable approach to extracting mobility patterns based on geographic routes from massive social media data. To support interactive visualization of mobility patterns by a large number of online users, this approach is implemented as a workflow of data processing, routing optimization, and multi-scale mapping. The scalability of the approach is demonstrated through both a suite of simulation experiments and a cyberGIS application of social media data.