Analysis of tens of thousands of comments from partisan online communities—coded with an AI-assisted method—showed more slippery-slope structures in conservative forums, and such comments tended to receive more approval from other users.