Research on the use of augmented reality (AR) in education has received a lot of attention in recent years. Based on many systematic reviews and meta-analyses, it has been concluded that AR is effective. Recently, however, researchers have criticized the fact that the empirical basis for this conclusion is based on results from methodologically problematic media comparison studies. However, an analysis of the literature and quantitative evidence for this claim are lacking. In this research project, this research gap was addressed using the Systematic Review method. A total of 92 primary studies from the top 12 Educational Technology journals were coded and analyzed. The results show that research on AR in education is based on media comparison studies: 80% of the studies compare AR to another medium or technology. Few studies examine how and when learning with AR is effective. In addition, results show that over the years, since 2009, more media comparison studies have been published than other research types. We summarize why media comparison studies are problematic and discuss directions for future research on AR in education. This research shifts from the question if AR can be used in instruction to the more important questions of how and when learning and teaching with AR works.