Public service media (PSM) plays a vital democratic role by offering unbiased content. The German Interstate Media Treaty establishes the societal task as PSM “shall serve education, information, advice, and entertainment” (content categories). This study examines how editorial (domain experts) and algorithmic recommendations (third-party provider) affect the distribution of content categories in clip impressions, clicks, and playbacks. A bias exists if recommendations lead to a concentration of content categories compared to clips available. Our observational study (2021 and 2023) on a German PSM video-on-demand platform reveals a bias towards entertainment in all recommendations. In our study, information (46%) and entertainment (32%) have the largest percentage of available clips. We find biases ranging from 46% in impressions for editorial recommendations up to 84% in clicks on algorithmic recommendations. Overall, editorial recommendations have considerably less bias than algorithmic recommendations. After this study, the PSM decided to build its own algorithmic recommender system.