This article explores how counterterrorism knowledge practices affect the groups they study. We argue that these practices typically construct terrorist groups as ontologically stable and organizationally rational, which makes them appear familiar to, and so governable by, counterterrorism organizations. We show that by excluding prevalent local knowledge, Western counterterrorism policy discourses assign the power to construct the category of “terrorist” to those without daily lived experience of the “terrorists” in question. This undermines different ways of knowing what sustains these groups, what might eradicate them and, more importantly, what might make their ability to pose a serious threat seem unlikely, or even absurd, to those whose support they purportedly need to survive as terrorists. Using evidence from Yemen, we show that groups labelled as “terrorists” do not fit into the stable categories that counterterrorism organizations require to produce actionable targets. We argue that while imposing such categories helps counterterrorists find targets that reflect their assumptions, it also generates pathways for violent actors to evolve and reproduce.