The model of descriptive representation, where representative assemblies mirror the social composition of electorates, is a salient issue in contemporary debates about the quality of representative government. In this vein, proponents stress its potentials to reconnect citizens and the state and to effectively enhance political trust in contemporary democracies. This paper contributes to this debate in theoretical and empirical ways. Theoretically, it argues that descriptive representation may facilitate political trust but that this is conditioned by voter characteristics and the behavior of representatives. Specifically, we argue that voters react to descriptive representation contingent upon their group identities and the level of substantive representation that descriptive representatives provide. Empirically, we explore this argument on the basis of an original conjoint experiment fielded in Germany, the UK and Switzerland. In this experiment, we randomize both the descriptive traits and policy positions of fictitious legislators to explore their direct effects on political trust and also how they interact in this regard. We combine our experimental set-up with a survey component to draw inferences about how voter characteristics condition the elite level effects on political trust. Our comparative design allows to explore the robustness of our results across different institutional contexts.