This paper studies a model of costly sequential search among risky alternatives performed by a group of agents. The learning process stops and the best uncovered option is implemented when the agents unanimously agree to stop, or when all the projects have been researched. Both the implemented project and all the information gathered during the search process are public goods. I show that the equilibrium path implements the same project based on the same information, gathered in the same order as the social planner. At the same time, due to free-riding, search in teams does lead to a delay at each stage of the learning process, the size of which grows with search costs. Consequently, the team manager prefers to delegate search to an individual agent, while every agent prefers searching with a partner, since in the latter case she collects the same reward, but only pays the search cost half the time.