An increasing fake paper problem is a cause for concern in the scientific community. These papers look scientific but contain manipulated data or are completely fictitious. So-called paper mills produce fake papers on a large scale and publish them in the name of people who buy authorship. The aim of this study was to learn more about the characteristics of fake papers at the metadata level. We also investigated whether some of these characteristics could be used to detect fake papers. For that purpose, we examined metadata of 12 fake papers that were retracted by Naunyn-Schmiedeberg's Archives of Pharmacology (NSAP) in recent years. We also compared many of these metadata with those of a reference group of 733 articles published by NSAP. It turned out that in many characteristics the fake papers we examined did not differ substantially from the other articles. It was only noticeable that the fake papers came almost exclusively from a certain country, used non-institutional email addresses more often than average, and referenced dubious literature significantly more often. However, these three features are only of limited use in identifying fake papers. We were also able to show that fake papers not only contaminate the scientific record while they are unidentified but also continue to do so even after retraction. Our results indicate that fake papers are well made and resemble honest papers even at the metadata level. Because they contaminate the scientific record in the long term and this cannot be fully contained even by their retraction, it is particularly important to identify them before publication. Further research on the topic of fake papers is therefore urgently needed.
