Recently, Casasanto and Lupyan (2015) have proposed an appealing and daring thesis: there are no context-independent concepts-that is, all concepts are ad hoc concepts. They argue that the seeming stability of concepts is merely due to commonalities across their different instantiations but that, in fact, there is nothing invariant in them. In their view, concepts only exist when they are instantiated for categorizing, communicating, drawing inferences, etc., and those instantiations are produced on the fly from a set of contextual cues. However, the main weakness of Casasanto and Lupyan’s framework is that it lacks a proposal for articulating it within a theory on the structure of concepts. My aim is to show that the ad hoc cognition framework can be characterized by means of a prototype theory of concepts developed in terms of a conceptual similarity space.