Using a two-level randomized experiment, we find that job search assistance (JSA) reduces unemployment among the treated but also creates displacement effects. Analyses of mechanisms show that vacancy referrals from caseworkers to job seekers explain the positive effects for the treated by helping the job seekers apply to the most relevant jobs earlier. We also find that the overall assessment of JSA hinges on how the displacement effects hit the labor market and to what extent firms react by opening new vacancies. The displacement is larger in weak labor markets, and we find no displacement of resources.