Rather than consider the direct impact of generative AI systems on art and authorship, this chapter explores how such technologies dictate the flow of information between authors and readers in online networks. The 2010s saw an increased spread of disinformation online resulting in political polarisation and disruption, notably influenced by the use of AI systems to moderate, recommend and personalise content for social media users. While ostensibly neutral authorities that function “for us” by recommending personalised content, these systems actually contributed to the post-truth confusion, division and mistrust of the 2010s that has only been exacerbated in the modern day. This chapter argues that personalised newsfeeds are a form of what philosopher Eugene Thacker terms weird media i.e., a mediation between the known human world and the unknown non-human world. Therefore, AI recommendation systems do not simply dictate flows of information but, rather, they construct coherent but entirely fictitious worlds for users. To fully appreciate how these systems influence the production and distribution of information, particularly in socio-political contexts, this chapter focuses on instances of apparently malfunctioning systems. By no longer functioning “for us”, these broken systems reveal the nonsensical and non-human nature of AI authorship.