Online labor platforms were expected to flatten labor markets by reducing the importance of worker location and, as a result, promote employment and wage growth in developing countries based on cost differentials. We test these propositions using transactional data from Nubelo, a large online labor platform for Spanish-speaking employers/freelance workers. The results suggest that information-related frictions long observed in traditional labor markets may be exacerbated in online platforms, resulting in worker discrimination based on country of origin. We show that, after controlling for observable workers’ characteristics and their job bids, foreign job-seekers are 42 percent less likely to win contracts from Spanish employers, which represent about two-thirds of all employers in the platform. We attribute this result to the activation of stereotypes that orient employers’ hiring decisions, in the absence of verifiable information about the quality of individual workers. We draw implications for platform design and the governance of online labor contracts.