As opposed to traditional forms of repression that were exercised almost exclusively within the jurisdiction of authoritarian states, the availability of social media has globalised both the expression of dissent by diaspora communities and the (transnational) reach of repressive actors. What makes a dissident’s online commentary dangerous enough to provoke prosecution or disappearance in an authoritarian regime? To answer this question, we investigate the online behaviour of critics and activists in repressive regimes from 2011 to 2024, examining what distinguishes those who are targeted by the state from those who remain unharmed. By focusing on what individuals post online and how they frame their comments, this study sheds light on the mechanisms by which authoritarian governments identify and suppress dissent. Our approach combines data from diverse digital sources, including social media posts, blogs, and public statements, to create a dataset of individuals who have faced state repression and those who have not. Using natural language processing (NLP) techniques, we systematically analyse the linguistic and contextual features of these online messages. The analysis highlights how certain expressions of dissent or criticism intersect with the strategies and priorities of authoritarian regimes, making some individuals more vulnerable to targeting. This study focuses on the methodological challenges of identifying repressive triggers in digital expression, providing a framework for future research on state surveillance and control. By examining the interaction between language and power, we offer insights into the risks associated with online dissent in authoritarian contexts.