Automated literature reviews have the potential to accelerate knowledge synthesis and provide new insights. We present a system that, starting with a set of “seed papers,” recommends related papers for a topic. To develop and evaluate these methods at a large scale, we use the reference lists from hundreds of existing review papers as labeled data, which can be used to train supervised classifiers. We find that our methods can identify many of the most relevant papers for a literature review from a large set of candidate papers, even when starting with a very small set of seed papers. This process can be adapted to identify previously undiscovered articles relevant to a topic.