There are some "generators" of scientific papers (including computer science, physics, mathematics or philosophy). We explain how these generators work and we present a method detecting generated texts. Until now, these false papers are fairly easy to identify because they are meaningless and mostly because they do not have the diversity of human vocabulary and syntax. However, several generated papers are appearing in scientific conferences and in the major bibliographic databases such as Scopus or WoK. Finally, there are several ways to improve these generators which may provide in the future, a useful help for text writing.