Abstract: This paper develops a model of political regime change and media censorship. An authoritarian regime can be overthrown if enough citizens join a coup. The citizens don't know the realized quality of the regime but receive correlated signals about it from media outlets. They also use social media to learn about the tactics to be used in an uprising. The regime can engage in censorship activities to reduce the probability of a coup. It might choose to censor media outlets which transmit to the citizens a signal about its quality (content censorship). This will make the regime seem to be of higher quality than it actually is. Alternatively, the regime might close down the channel through which the citizens learn about the protest tactic (coordination censorship). This will make the citizens' coordination harder to sustain. Finally, the regime might adapt both types of censorship, i.e., content censorship and coordination censorship. We characterize equilibrium of this coordination game and provide equilibrium comparative statics. Our results suggest that there is an inverse U-shaped relationship between the regime's type and the adopted censorship type. The more competent the regime, the more likely it will imply either type of censorship up to some point. Then, the less likely it will censor. Moreover, a regime applying coordination censorship is likely to be less competent than a regime applying content censorship.