There are many books and courses about how UML should be used, but how is UML used in the real world? This thesis attempts to answer this question by downloading and analyzing 522 UML models from the Internet. The analysis focuses on quantitatively measurable model parameters to create an usage overview over all models. This way it could be observed that most models are of rather small size and only use few language aspects UML has to offer. The usage is distributed rather unequal, with Class and Use Case being the most popular Language Units of UML. Also almost every Language Unit has a few concepts used most of the time and a few concepts used practically never. Based on this data it is further discussed, how and why people might create models the way they do and how this affects UML.