Health risk behavior in student populations is an issue of major concern, and students' risk levels are difficult to determine. In this study, we explore the extent to which information disclosed publicly on Facebook provides reliable indications of five real-life health behaviors. Questionnaire data and Facebook contents (2928 items) on alcohol use, smoking, illicit drug use, (un)healthy nutrition, and participation in sports of 71 respondents were collected and analyzed. The study shows that one can analyze Facebook profiles to reliably associate profile owners' smoking and sport behavior. It also shows that regarding alcohol use, some Facebook profile elements are indicative of real-life drinking. We discuss and suggest improved methods of coding disclosed public data, which may lead to reliable indications of peoples' real-life behavior. © 2014 Elsevier Ltd. All rights reserved.