This paper analyzes the design of optimal unemployment insurance in a search equilibrium framework where search effort among the unemployed is not perfectly observable. We examine to what extent the optimal policy involves monitoring of search effort and benefit sanctions if observed search is deemed insufficient. We find that introducing monitoring and sanctions represents a welfare improvement for reasonable estimates of monitoring costs; this conclusion holds both relative to a system featuring indefinite payments of benefits and a system with a time limit on unemployment benefit receipt. The optimal sanction rates implied by our calibrated model are much higher than the sanction rates typically observed in European labor markets.