This work examines the benefits and risks of using available classes of uncertainty indexes for policy purposes, clustered in three broad categories: survey-based, model-based, and news-based. In both policy discussions and the academic literature newsbased indexes are the ones that have recently gained the most attention. We argue that the reasons behind this are their intuitiveness, transparency and real-time characteristics. The main trouble with these indexes, as they are constructed today, is their noisiness. We then suggest that, for policy purposes, it would be better to disregard very high frequency movements in the series. Finally, we highlight that well-developed probabilistic surveys still represent a hard-to-beat benchmark when one is interested in uncertainty concerning specific variables as opposed to more abstract concepts such as Economic Policy Uncertainty.