identifying a set of seed content items, wherein each seed content item in the set of seed content items corresponds to a violative content item;

determining, using playback histories indicating consumption of content items, connections between a first content item and a first audience that has consumed the first content item and that have consumed at least a threshold number of seed content items from the set of seed content items;

providing information corresponding to the connections as an input to a machine learning model;

receiving, as an output from the machine learning model, likelihoods that respective content items are violative content items;

storing a set of content items, selected using the output from the machine learning model, as candidate content items in accordance with a determination that each of the set of content items satisfies likelihood criteria; and

in accordance with a determination that a particular content item of the candidate content items satisfies the likelihood criteria, displaying a user interface element that includes a warning that the particular content item is likely to be a violative content item.