Political parties, social movements, and hundreds of millions of people now use social media to engage in political activity and discussion. The goal of this dissertation is to understand how different actors use these technologies to organize and to advance their political claims. The first empirical chapter studies the far-right group Britain First and how it used Facebook to reach more than two million people. I collect data on the daily changes in the group’s activity, support, and recruitment over a three year period and analyze it using time series regression. I find that these processes are largely explained by prior social media activity, drawing attention to how the group exploited algorithmic filtering to its advantage. The next chapter focuses on the role of opinion leaders in online political discussions preceding the Brexit referendum. I construct a panel dataset containing over one hundred thousand Facebook users active on the pages of political parties. Potential influentials are identified and the content of their opinions is inferred using a machine-learning classifier. I examine the effect of exposure to their comments on users’ subsequent interactions with Brexit campaign pages using a quasi-experimental technique. I find that pro-Leave opinion leaders dominated the online conversation and that exposure to their comments is associated with changes in other users’ online behavior but not their opinion expressions. In a third empirical chapter, I develop an agent-based model to reexamine the theory of opinion leadership. I show how the effect of opinion leadership is sensitive to the prior distribution of opinions in the population, as well as the number of opinion leaders and the density of the network. I conclude by discussing the substantive, theoretical, and methodological contributions of the dissertation, as well as opportunities for further study.