Semantic verbal fluency (SVF), a psychological assessment method used in experimental research and clinical practice, requires participants to produce as many words as possible from a given superordinate category (e.g., “animals,” “vehicles”). Features of responses, such as the prototypicality and ordering of items, are then interpreted as if revealing details about the organisation—or, in instances of ostensibly atypical performance, disorganisation—of participants’ underlying conceptual and/or semantic systems. In this paper, we draw upon perspectives from Discursive Psychology, particularly the work of Derek Edwards (e.g., Edwards, 1997), to argue against this position. Following critical discussion of SVF’s strongly cognitivist theoretical foundations, we present analyses of social interactions across various contexts, including the real-life administration of the paradigm with a child with autism, to suggest that performance is unavoidably socially mediated rather than solely internally driven. Our arguments challenge SVF’s validity and its role in the description of “cognitive disorder.”