YouTube Networks Use Real Actors with AI Scripts to Evade Platform Labels

Researchers at Riddance AI recently discovered a network of around a dozen YouTube channels utilizing real actors alongside AI-generated scripts. [1] The company, founded by Jeremy Carrasco and Mason Broxham, found that these channels paired human presenters with artificial intelligence in an effort to escape YouTube's AI labels. [1] Since the spring, this network of videos has accumulated 45 million views and over 90,000 comments by publishing nearly identical straight-to-camera monologues every few days. [1]

This discovery follows an early August report detailing a flood of AI-assisted political content that spread false claims on the platform. [1] Earlier videos featured AI avatars spreading misinformation, such as false assertions that Kroger, Costco, and Shell were closing California locations. [1] After the publication of that initial story, YouTube removed several of the offending channels for content violations. [1]

The transition from fully synthetic videos to hybrid models utilizing real actors indicates that content creators are actively adapting their strategies to bypass platform moderation. While YouTube successfully removed several channels relying on AI avatars to spread false claims about retail closures, this new network demonstrates that pairing human faces with AI-generated scripts can effectively evade automated labels while still achieving massive reach. This marks a clear evolution in evasion tactics, shifting the burden of detection from visual artifacts to narrative generation. What the sources don't address: whether YouTube plans to update its moderation policies or detection tools to specifically target these hybrid content networks.