In this work, I develop an expressivist approach to the analysis of knowledge ascriptions. Expressivism is a philosophical position that attempts to explain the meaning of various kinds of natural language expressions, rejecting that their communicative function is to describe how the world is. I wish to apply this form of analysis to a certain family of expressions, namely, those we use to ascribe knowledge. Thus, epistemic expressivism holds that knowledge ascriptions do not express factual relations between a subject and some propositional content. Rather, we use them to express our acceptance of systems of epistemic norms that change from context to context. In this work, I scrutinize two recent attempts (Chrisman 2007, Field 2009) at applying this mode of analysis to the attribution of knowledge. I argue that epistemic expressivism is more successful than its theoretical competitors –such as epistemic contextualism or invariantism– in explaining a certain array of phenomena. However, it also faces a number of problems, some of which constitute important challenges for its position. Finally, I argue that the expressivist approach conveys a sensibly relativistic picture of knowledge, which reflects our epistemic practices more accurately and which can also be positively applied in order to enrich our democratic practices.