This article critically assesses Linda Zagzebski’s use of direct reference theory within her agent‑based and exemplarist virtue ethics and argues that importing this semantic framework from the philosophy of language into moral philosophy generates significant theoretical and practical difficulties. It is shown that, in Zagzebski’s system, direct reference does not play a foundational or constitutive role in explaining the nature of virtue; rather, it serves primarily an educational and heuristic function in transmitting moral concepts to a general audience, since she ultimately refers the precise understanding of those concepts to experts. Moreover, the theory faces a paradox in the identification of moral exemplars: while virtue originates from the internal motivations of the moral agent, its recognition and fixation are entrusted either to common moral opinion or to specialists who are not necessarily virtuous themselves. Furthermore, grounding moral knowledge in particular individuals situated in specific spatiotemporal contexts risks fostering moral authoritarianism and reducing universal virtue to limited instances. The article concludes that applying direct reference undermines, rather than strengthens, the conceptual coherence of exemplarism, though its true value may lie in its educational, formative, and cultural capacity. Finally, a revisionary proposal based on a graded (tashkīkī) interpretation of virtue and exemplars is suggested.