Coordinated Inauthentic Behavior on TikTok: Challenges and Opportunities for Detection in a Video-First Ecosystem

TL;DR

This paper develops a novel, network-based framework for detecting coordinated inauthentic behavior (CIB) on TikTok, a video-first platform with multimodal content. It introduces seven behavioral traces and a TF-IDF-based similarity pipeline, augmented by two pruning strategies, to identify dense coordination subgraphs in a large election-related dataset. Applying the framework to 1.35M TikTok videos from the 2024 U.S. election reveals multiple CIB patterns, including synchronized posting, identical hashtag sequences, and AI-generated voiceovers, while showing that TikTok-native signals like Duet/Stitch may be less informative for coordination. The work highlights platform-specific challenges, including API limitations and enforcement gaps, and argues for multimodal, trace-fused detection approaches to robustly counter short-form video manipulation on social platforms.

Abstract

Detecting coordinated inauthentic behavior (CIB) is central to the study of online influence operations. However, most methods focus on text-centric platforms, leaving video-first ecosystems like TikTok largely unexplored. To address this gap, we develop and evaluate a computational framework for detecting CIB on TikTok, leveraging a network-based approach adapted to the platform's unique content and interaction structures. Building on existing approaches, we construct user similarity networks based on shared behaviors, including synchronized posting, repeated use of similar captions, multimedia content reuse, and hashtag sequence overlap, and apply graph pruning techniques to identify dense networks of likely coordinated accounts. Analyzing a dataset of 793K TikTok videos related to the 2024 U.S. Presidential Election, we uncover a range of coordinated activities, from synchronized amplification of political narratives to semi-automated content replication using AI-generated voiceovers and split-screen video formats. Our findings show that while traditional coordination indicators generalize well to TikTok, other signals, such as those based on textual similarity of video transcripts or Duet and Stitch interactions, prove ineffective, highlighting the platform's distinct content norms and interaction mechanics. This work provides the first empirical foundation for studying and detecting CIB on TikTok, paving the way for future research into influence operations in short-form video platforms.

Figures (19)

• Figure 1: Cluster of coordinated accounts sharing identical hashtag sequences. Usernames within the cluster show strong similarity, all beginning with a prefix linked to a U.S. presidential candidate, denoted here as "CandidateX" to preserve anonymity and privacy. To reduce visual clutter, we display only the usernames of accounts that posted more than five videos with AI-generated voiceovers.
• Figure 2: Synchronous posting times of coordinated accounts sharing the same hashtag sequences.
• Figure 3: Examples of videos shared by coordinated accounts sharing the same hashtag sequences. The watermark in the top-left corner and the description at the bottom were obscured to preserve anonymity and privacy.
• Figure 4: Clustering of spectrogram vectors extracted from videos shared by coordinated accounts using identical hashtag sequences. The inset shows a t-SNE dimensionality reduction of the spectrograms, highlighting distinct clusters corresponding to different AI-generated synthetic voices.
• Figure 5: Examples of split-screen videos and profiles from coordinated accounts sharing identical hashtag sequences.