Passenger flow data are necessary for making and coordinating operational plans for urban rail transit (URT) systems; the availability and the service state of those systems directly influence the activity of a city and its people. Although many transit assignment models have been developed, the results of passenger flows estimated by these models as well as assumptions made in the estimation process, especially for large-scale, complex, and dynamically changing URT networks, had not been validated. This paper proposes a methodology that can validate existing URT assignment models by using automatic fare collection data and a cluster analysis technique. Initial applications to the URT system of Shanghai, China, which is one of the largest in the world, show that the proposed approach works well and can efficiently find the origin–destination pairs in which passengers' route choices are misestimated by those assignment models. The analysis suggests that several factors result in errors (for the URT assignment model used in Shanghai). These factors include the threshold for the difference in travel costs, a misrepresentation of the transferring cost, and inadequate values for the standard deviation. This information is useful for detecting errors in existing URT assignment models, leading to improvements.