We study a search model with incomplete information in which a decision maker has to select one agent from a sequence. The value of the agents for the principal is their private information, which can be reported to the principal. The first-order issue for the principal is the tradeoff between the quality of the selected agent and the ability to infer the agents' quality from their reports: selecting the agent with the highest reported quality destroys incentives for truthful reporting. The optimal policy balances these considerations by selecting an agent with a lower reported quality with positive probability and restricting the number of agents that are sampled. The model is tractable and we obtain an interesting non-stationary optimal search policy, which stands in stark contrast to the optimal cutoff policy under complete information.