Prompt, Upload, Repeat: Agentic AI Accounts Flood TikTok

Assessing the impact of Agentic AI Accounts on the proliferation of harmful AI slop

03-12-2025

The surge of AI slop, deceptive synthetic videos and images flooding our feeds, is driven in large part by Agentic AI Accounts (AAAs). In this third report in our series, we examine how these accounts operate. Their goal is to “game the algorithm” through sheer volume, using tools and websites to refine content and rapidly test platform systems at scale.

For more on AAAs, see our AAA Codebook.

We analyzed 354 such accounts over one month across 20 languages.

Key findings include:

Over 43,000 mostly AI-made posts generated 4.5 billion views
More than 65% of accounts were created in early 2025
TikTok labels less than 1.38% of this content as AI-generated
Nearly ⅓ of accounts and ½ of the top 10 showed content sexualizing female bodies, including minors
False news, anti-immigrant narratives, and explicit trends were widespread
55% of AI content remains unlabelled; only 10% of creators label consistently

The investigation shows how automated accounts now compete directly with human creators for visibility and influence.

Solution & Methodology

In AI Forensics’ recent investigation on the use of AI imagery on Instagram and TikTok, we introduced the term Agentic AI Accounts (AAAs). These accounts automate content creation, facilitating rapid, repetitive testing of platform algorithms. To identify relevant AAAs, we turned to TikTok’s “For You feed” (FYF). We then combined snowball sampling with the creation of a personalized account. To make an initial distinction between AI and non-AI content, we relied on The Human Guide to Detecting AI Imagery.

03-12-2025
The Guardian covers AI Forensic's report, including 354 AI-focused accounts with 4.5 billion views. It highlights that, among this content, is anti-immigrant material.
03-12-2025
Der Spiegel covered the report, highlighting that videos shared by Agentic AI Accounts often feature girls posing suggestively, garnering millions of views.
04-12-2025
Le Parisien reported on our research, highlighting the range of harmful AI content proliferating on TikTok and noting our creation of the term "Agentic AI Accounts."