Online dating sites have become popular platforms for people to look for romantic partners, providing an unprecedented level of access to potential dates that is otherwise not available through traditional means. Characterization of the user online dating behavior helps us to obtain a deep understanding of their dating preference and make better recommendations on potential dates. In this paper we study the user online dating behavior and preference using a large real-world dataset from a major online dating site in China. In particular, we characterize the temporal behavior, message send and reply behavior of users, study how users online dating behaviors correlate with various user attributes, and investigate how users’ actual online dating behaviors deviate from their stated preferences. Our results show that on average a male sends out more messages but receives fewer messages than a female. A female is more likely to be contacted but less likely to reply to a message than a male. The number of messages that a user sends out and receives per week quickly decreases with time, especially for female users. Most messages are replied to within a short time frame with a median delay of around 9 h. Many of the user messaging behaviors align with notions in social and evolutionary psychology: males tend to look for younger females while females place more emphasis on the socioeconomic status (e.g., income, education level) of a potential date. The geographic distance between two users and the photo count of users play an important role in their dating behavior. We show that it is important to differentiate between users’ true preferences and random selection. Some user behaviors in choosing attributes in a potential date may largely be a result of random selection. We also find that while both males and females are more likely to reply to users whose attributes come closest to the stated preferences of the receivers, there is significant discrepancy between a user’s stated dating preference and his/her actual online dating behavior. We further characterize how users actual dating behavior deviate from their stated preference. These results can provide valuable guidelines to the design of a recommendation engine for potential dates.