The large volume of published scientific literature, combined with concerns about the quality of scientific research itself, presents unique challenges for researchers. Specifically, it is increasingly difficult to identify research that has later been replicated – or, in some cases, been the subject of a failed replication attempt. scite.ai is a tool that addresses this issue by using machine learning to extract the text of citations in published articles and identifying those which support, contrast, or simply mention the findings presented in the paper they are citing. So far, scite has extracted over 1.2 billion citations from over 33 million published articles. This talk will explore how scite can be used in a variety of common scientific tasks (orienting oneself to a new subfield, preparing a grant application, etc.), as well as discuss the broader implications of applying machine learning to metascientific questions.