Audit experiments on TikTok show asymmetric partisan exposure that is not explained by observable engagement metrics, with Republican-leaning accounts receiving more aligned content and Democratic-leaning accounts more cross-partisan recommendations.
