This paper builds on the existing work of United Nations bodies to provide more granular and well-calibrated guidance on the application of these provisions to online hate speech, particularly in fragile settings. First, it clarifies why international human rights law matters for states, companies and civil society organisations operating in this space. Second, it provides a taxonomy of online hate speech, based on the distinct legal consequences that different forms of hate speech attract under Articles 19 and 20 of the ICCPR. Three categories are proposed, along with respective content moderation measures: a) prohibited, b) limited, and c) free speech acts. The paper then applies this framework to fragile settings.