Within the last decade, social network sites (SNSs) have come to play an increasingly important role in relation to both everyday and political life in Europe and North America. This gives rise to new forms of cultural and political participation, but also new modalities of manipulation. This paper explores how disguised propaganda on SNSs can be studied methodologically and analytically by combining multi-sited online ethnography (Hine, 2015; Marcus, 1998) and social media critical discourse studies (SM-CDS) (Khosravinik, 2017; Unger, Wodak, & KhosraviNik, 2016; Reisigl & Wodak, 2001). As numerous scholars have argued, SNSs should not simply be viewed as transparent or static platforms on which sociality unfolds (Langlois & Elmer, 2013; van Dijck, 2013). Rather, they represent contingent online spaces continuously shaped through the intermesh between social and technological processes. Political discourses on SNSs consequently require researchers to critically examine, not only how discourses are constructed within texts, but also how they arise through socio-technical discursive practices and processes. Based on a six-month multi-sited online ethnography, the paper discusses how participant-observational fieldwork and SM-CDS can be applied in conjunction to study the (re-)production and proliferation of racist discourses through fake SNS profiles. The study revolves around 11 Danish Facebook pages using fake Muslim identities to provoke Danish Facebook users. According to the pages, Muslims in Denmark are part of a conspiracy to take over the country, killing and raping non-Muslim Danes in the process (Farkas et al., 2017). In total, the pages received more than 20,000 comments, a majority of which expressed belief in the proclaimed authorship and aggression towards the pages and Muslims in general. The paper explores the socio-technical construction, maintenance, negotiation and contestation of racist discourses on these Facebook pages. Additionally, the paper discusses the methodological and analytical challenges of studying this type of online phenomenon. The paper concludes by arguing that multi-sited online ethnography can complement and enhance SM-CDS, although more scholarly work is urgently needed on this topic.