We investigate the role of information frictions in migration decisions. We show that even without data on what migrants know, most migration models are partially identified under rich patterns of incomplete information. We develop novel moment inequalities estimators for key preference parameters that allow workers' information sets, migration costs, amenities, and price levels to vary flexibly and remain unobserved by the researcher. Applying our estimators to internal migration flows in Brazil, we find that prior work underestimates the importance of expected wages in workers' migration choices and that workers face substantial information frictions. Workers in regions with better internet access and higher population density have more precise wage information, especially about wages in nearby areas. We quantify how information frictions alter migration flows, welfare, and the predicted effect of reducing migration costs.