For some years now, news sites around the world are increasingly confronted with abusive user comments in their respective comment sections and discussion forums. While these spaces were long seen as promising instruments of democratic participation, they now have a reputation as spaces full of insults and hate speech. Since this not only poses a threat to social cohesion but can also compromise the image of a news site, many platforms have taken measures to regulate the comments on their sites since then. Some have published community guidelines, hired moderation teams and implemented monitoring software. As an additional measure, many adapted the technological design and the features of their comment spaces to gain more control over the posted comments. This includes, for example, requiring commenters to register with the site, sorting of comment threads and various degrees of anonymization. Many authors refer to this technological design of comment spaces as "discourse architecture." The theoretical argument behind this term is that the way comment spaces are "built" influences how commenters behave within them. This perspective is particularly interesting from the point of view of journalism research, since the relationship between editorial staff and audience is manifested in such technological architectures. Several studies have analyzed and compared various discourse architectures in order to investigate possible effects on commenting behavior. However, there is still a lack of a systematic analysis in this field. Apart from individual case studies, there are no findings on the diversity of discourse architectures which provide information on the technical conditions of audience participation on the Internet. On the theoretical basis of the discourse architecture approach, this study investigates two research questions: How are the included discourse architectures designed (RQ1)? And what types of discourse architectures can we identify (RQ2)? In order to answer these questions, we conducted a standardized analysis of 361 German news sites, which produced three key findings. Firstly, with regard to RQ1, we found that 173 of these 361 news sites offer comments sections, whereas only 24 offer discussion forums. In contrast, almost all sites in the sample have an additional Facebook page. Al-though we have not checked whether these pages actually contain posts and comments, against this back-ground we can nevertheless assume that the discourse architecture of Facebook has become the most important technological infrastructure for commenting news articles in Germany. Acknowledging the low deliberative quality of user discussions on Facebook revealed by earlier studies, this would be quite problematic with regard to social integration. Secondly, the detailed analysis of the comment sections showed that most news sites do not exhaust the possibilities of using technical discourse architectures to gain more control over the discussions of users and users. Overall, the technological design of the comment sections is quite inclusive, not very restrictive and only weakly regulated. The most popular features are required registration, rating of comments, opprtunities to report comments and the restriction of comment sections to certain topics. Thirdly, with regard to RQ2, five distinct types of discourse architectures for comment sections could be identified within the sample. They differ in terms of their combinations of features and as well as in terms of their outreach. Additionally, we found a significant correlation between the outreach of the news sites and the number of features that strengthen editorial control over the comments.