Abstract: One major motivation for conceptual engineering is that some ways in which we think and talk about reality are defective. Mona Simion, however, argues that it is permissible to engineer non-defective concepts too even when they are good enough. Still, Simion holds that there needs to be normative constraints on when this is legitimate: Engineering mustn’t come with an epistemic loss. I don’t dispute that we can have reasons to engage in continual conceptual maintenance. But some conceptual resources should prima facie be fixed because of serious defects involved: ones I term ‘ideologically defective concepts.’ Contra Simion, it isn’t possible to engineer such concepts without an epistemic loss and this is the normative ground to engineer. I am not alone in thinking so. Paul-Mikhail Catapang Podosky argues for an expanded version of Simion’s view precisely since Simion’s constraint would make engineering ideologically defective concepts impermissible. Podosky, I contend, is right about ways in which Simion’s normative constraint falls short. However, I argue, his expanded version also falls short: Podosky’s understanding of how ideological concepts work is not apt to spell out what it means for conceptual defects to hinge on ideology. Here I advance an alternative diagnosis of such defectiveness. My analysis suggests that it is extremely difficult, if not impossible, to re-engineer ideological concepts. But the engineer can still engage in de novo engineering together with conceptual ethics. This makes conceptual engineering of ideological concepts much trickier than Simion and Podosky’s views suggest. But ethics is hard and conceptual ethics is no exception.