Recent discussions of cognitive enhancement often note that drugs and technologies that improve cognitive performance may do so at the risk of “cheapening” our resulting cognitive achievements (e.g., Kass, Life, liberty and the defense of dignity: the challenge for bioethics, Encounter Books, San Francisco, 2004; Agar, Humanity’s end: why we should reject radical enhancement, MIT Press, Cambridge, 2010; Sandel, The case against perfection. Harvard University Press, Cambridge, 2007; Sandel, The case against perfection: what’s wrong with designer children, bionic athletes, and genetic engineering?”. In: Holland (ed) Arguing about bioethics, Routledge, London, 2012; Harris in Bioethics 25:102–111, 2011). While there are several possible responses to this worry, we will highlight what we take to be one of the most promising—one which draws on a recent strand of thinking in social and virtue epistemology to construct an integrationist defence of cognitive enhancement. (e.g., Pritchard in Synthese 175:133–151, 2010; Palermos in Synthese 192:2955–2286, 2015; Clark in Synthese 192:3757–3375, 2015). According to such a line, there is—despite initial appearances to the contrary—no genuine tension between using enhancements to attain our goals and achieving these goals in a valuable way provided the relevant enhancement is appropriately integrated into the agent’s cognitive architecture (in some suitably specified way). In this paper, however, we show that the kind of integration recommended by such views will likely come at a high cost. More specifically, we highlight a dilemma for users of pharmacological cognitive enhancement: they can (1) meet the conditions for cognitive integration (and on this basis attain valuable achievements) at the significant risk of dangerous dependency, or (2) remain free of such dependency while foregoing integration and the valuable achievements that such integration enables. After motivating and clarifying the import of this dilemma, we offer recommendations for how future cognitive enhancement research may offer potential routes for navigating past it.