While online dating platforms offer new IT-enabled capabilities which do not exist in the physical world before, little is known about whether any fundamental matching rules are reshaped in the online environment. In this paper, we address the gap by studying one such factor, i.e. mate physical attractiveness, in an online dating platform. By using a unique dataset and machine-learning based algorithmic approach, the study successfully overcomes various confounding issues, selection bias and physical attractiveness measurement issues and estimates the physical attractiveness effect in online users’ dating decision. Results reveal the essence of physical attractiveness in online context and the disappearing geographic boundary. The findings and methods are essential to both our understanding of the mechanisms that drive match mating online and our knowledge of how to propagate them in various fields where large scales of objective physical attractiveness and behavioral data are emerging.