Tiktok's Research API: Problems without Explanations

Our investigation reveals TikTok's Research API fails to provide metadata for one in eight videos, undermining independent research efforts.

12-06-2025

Project overview

Report cover

Following the Digital Services Act requirements for platform transparency, TikTok expanded its Research API access in Europe in July 2023. However, our comprehensive investigation reveals significant limitations that undermine the API's reliability for independent research:

  • One in eight videos from data donations cannot be retrieved through the API
  • Official TikTok company videos, including CEO statements with 30+ million views, are inaccessible
  • Prominent accounts like Taylor Swift, Brook Monk, and major news outlets have content inaccessible
  • Thousands of advertisements remain hidden despite their public availability
  • Random account exclusions affect approximately 1% of creators without explanation

This creates a fundamental reliability problem for researchers conducting platform accountability studies, algorithm audits, and content analysis using data donation methodologies.

We created a dashboard where we daily check a set of inaccessible videos to verify if the problem persists : https://playground.tiktok-audit.com/api-na/

Project Details

Solution & Methodology

Our investigation employed a twofold experiment: 1) Testing the API using data donation methodology with 260,000 TikTok URLs, systematically checking which videos could not retrieve metadata. When initial batch requests of 100 videos failed, we individually queried each ID over 64 days due to TikTok's 1,000 daily API call limit. 2) Conducting real-time monitoring by collecting 100 videos daily from German For You Pages for one week, then testing API availability for creators' recent content. We verified all findings through web scraping to confirm public availability.

Implications for Independent Research

The API's unreliability fundamentally compromises research validity, particularly for studies using data donation methodologies that rely on complete datasets. When 1/8 videos lacks metadata, researchers cannot accurately analyze user consumption patterns, content recommendation algorithms, or platform behavior. This is especially problematic for reactive investigations into issues like misinformation or harmful content trends where TikTok's 48-hour delay already limits timely analysis.

AI Forensics is proud to be trusted by