PoliTok-DE dataset reveals 13% of German political TikToks are deleted

A new research dataset called PoliTok-DE gives researchers an unprecedented look at what gets deleted on TikTok during German elections. Created by Tomas Ruiz, Andreas Nanz, and four co-authors from Munich universities, the corpus contains over 930,000 TikTok posts collected via the platform's research API and supplemented with web scraping for full multimodal media (video, audio, images, and text). The data covers two periods: the 2024 Saxony state election and the 2025 German federal election. Of those posts, more than 330,000 were later deleted from the platform — 18.7% from Saxony and 39.7% from the federal election. About two thirds of federal-election deletions were creator withdrawals, but the platform-deletion rate (13% of all posts) is 14 to 19 times higher than the rate TikTok officially reports. This massive discrepancy suggests TikTok's public figures on content removal may significantly undercount political content.

The dataset also includes a case study on an annotated subset of deleted posts, examining the intersection of intolerance and entertainment. The findings are striking: roughly one in five deleted posts conveyed intolerance, while a majority employed humor. PoliTok-DE is designed to advance two research agendas: substantive work on political communication, polarization, and hate speech, and methodological work on platform policies around deleted content and qualitative-quantitative multimodal analysis. By combining video, audio, text, and metadata with deletion tracking, it provides a rare empirical baseline for understanding how platforms like TikTok handle political speech during high-stakes elections — and how creators either self-censor or are removed by the platform.