This study investigates Amazon’s book recommendation system, uncovering cohesive communities of semantically similar books. The confinement within communities is extremely high, a user following Amazon’s recommendations needs tens of successive clicks to navigate away. We identify a large community of recommended books endorsing climate denialism, COVID-19 conspiracy theories, and advocating conservative views on social and gender issues. Performing a collaborative filtering analysis, relying on Amazon users reviews, reveals that books reviewed by the same users tend to be co-recommended by Amazon. This study not only contributes to addressing a gap in the literature by examining Amazon’s recommender systems, but also highlights that even non-personalized recommender systems may pose systemic risks by suggesting content with foreseeable negative effects on public health and civic discourse.