Large-scale TikTok audit finds no evidence of political shadow banning

A new study from arXiv (July 2026) tackles persistent allegations that TikTok shadow bans political content. Researchers collected a dense hourly panel of 556,946 follower-normalized views across 2,753 videos from 67 accounts curated into pro and anti stances on three contested topics: U.S. immigration enforcement, Trump coverage, and Israel/Palestine. A multi-step classifier identified on-topic videos, and stance was assigned from each account's curated side. The conventional pooled analysis appeared to confirm suppression, with a topic-conditional reach gap reaching p<10^-140. However, analyzing at the account level—the proper independent unit—every reach contrast became null after correction (BH-FDR q near 0.9). The study finds no evidence of moderate-to-large reach suppression on any topic. The design detected a clear asymmetry on engagement: oppositional content (anti-Trump, pro-Palestine) earned more engagement per view (Cliff's delta = -0.51 and -0.64; q<0.03), showing the framework can detect effects of that magnitude.

The apparent reach gap is an artifact of two factors: pseudoreplication (treating autocorrelated video-hours as independent observations) and confounding (the side that looked suppressed was larger and, on Israel/Palestine, posted mostly in Arabic). In this corpus, what appears to be a shadow ban is better explained by a more engaged audience than by a suppressed one. The authors conclude with requirements for credible visibility audits, emphasizing proper statistical units and confound control. The paper has implications for platform accountability, advertiser trust, and regulatory decisions built on the shadow ban narrative.