Exclusive: Paid actors, AI writing: How a new kind of…

In early August 2024, an investigation uncovered a sophisticated network of YouTube channels leveraging a hybrid approach to distribute political content. By combining AI-generated or enhanced scripts with real human actors, the operators of these channels sought to circumvent automated detection systems that typically flag purely AI-created content. Researchers at Riddance AI identified a cluster of approximately a dozen channels that amassed 45 million views and over 90,000 comments by recycling specific scripts, music, and formulas to target Democratic politicians with sensationalized or false economic and corruption claims.

What happened

The operation utilized a coordinated strategy to maintain a veneer of authenticity. Instead of relying on AI avatars, the network hired real human actors—identified by researchers as professional spokespeople available on platforms like Fiverr—to perform straight-to-camera monologues. These actors, such as Kareem M. Maize and Frank George, performed scripts that often featured identical phrasing, such as "here’s the part that should stop you cold," and reused specific b-minor musical scores.

The content focused on prominent Democratic mayors, governors, and members of Congress, utilizing aggressive, all-caps titles designed to lure viewers with exaggerated claims. Despite the professional presentation, the videos often contained false information, such as fabricated claims about corporate store closures or misrepresentations of legislative proposals. The network displayed clear signs of centralized coordination: channels frequently swapped personas, cross-linked to each other's profiles, and redirected viewers to shared domains. In one instance, an actor who previously fronted a channel criticizing Donald Trump reappeared on a pro-Trump channel after a period of dormancy, adopting a new persona to criticize progressives. This hybrid model—merging AI-assisted scripting with human performance—was specifically designed to evade YouTube’s AI labeling systems, allowing the network to gain significant traction before being identified by researchers and subsequently facing content violations.