By using social randomized experiments, treatment effects of active labor market policies can be evaluated under the stable unit treatment value assumption (SUTVA). This assumption will be violated in labor markets if treatment externalities exist (congestion or general equilibrium effects). We estimate the magnitude of those externalities using a large data set from a randomized experiment in Denmark. By comparing the control group in the treatment region with workers in unaffected regions we find some evidence that the job finding rate of the control group workers in the treatment region declines due to the treatment. Furthermore, we find a small increase in vacancy supply in the treatment region. In order to estimate the effects of implementing the program country wide, we construct an equilibrium search model with job search assistance and use our estimates to calibrate it. We find that although the negative congestion effects are substantial, an increase in the treatment rate does decrease unemployment.