The aim of this paper is twofold. First, I want to re-examine the epistemic status of elaborated delusions. Bortolotti (2016, 2020) claims that they can be epistemically innocent. However, I will show that this type of delusions is more unique than suggested by the existing analyses of their epistemic status. They typically cause more profound harms than other kinds of delusions, and in most cases, it would be counterproductive to classify them as epistemically beneficial or innocent. I will employ predictive (...) and phenomenological models of delusions and the enactivist notion of sensus communis to explain the harms and possible benefits of elaborated delusions and why I think the existing definition does not fully grasp them. Based on this analysis, I will propose changes to the conditions for epistemic innocence. (shrink)