AI - Generated “Cheapfakes” on YouTube: A Deceptive Content Phenomenon
Introduction
Mark Wahlberg's purported ousting from “The View” after a heated showdown with Joy Behar seemed like a blockbuster moment in daytime television. However, this was no real - life drama. Since 2015, Wahlberg hasn't been a guest on the show. This incident played out in a 25 - minute fan - fiction - style video, crafted using artificial intelligence, potentially deceiving 460,000 viewers eager for drama.
The Nature of “Cheapfakes”
Composition and Execution
Simon Clark, a cognitive psychologist at the University of Bristol specializing in AI - generated misinformation, describes these videos as “cheapfakes.” They are cobbled together from a hodgepodge of real images and video clips, accompanied by a basic AI voice - over and subtitles. For instance, the “Mark Wahlberg” video is mostly a still image with a grainy filter, relying solely on an AI - narrated, LLM - written script filled with clichés like “fist - clenching” and “jaw - wobbling” to entertain its audience.
Despite their shoddy production, the channel “Talk Show Gold” hosting this video has amassed over 88,000 subscribers. Many viewers are left in disbelief when informed that they've been watching “fake news.”
Psychological Manipulation
These videos employ rhetorical techniques that appeal to emotions, encouraging viewers to abandon critical thinking. By triggering moral outrage, they make it easier for audiences to engage with and spread misinformation. As Clark notes, outrage is a powerful emotion for driving engagement. When viewers feel outraged, they are more likely to share the video with like - minded friends and write long rants in the comments, regardless of whether the events are real or labeled as “AI - generated.”
YouTube Channels' Tactics
Camouflage and Deception
A WIRED investigation uncovered 120 YouTube channels using similar tactics. With misleading names such as Starfame, Media Buzz, and Celebrity Scoop, they blend in with real compilation clips from shows like “Jimmy Kimmel Live!” and “Today With Jenna & Friends” to gain credibility.
Their channel descriptions create the illusion of sensational tabloid outlets. Some bury AI - disclaimers under walls of text highlighting “all the best highlights” or “the most unforgettable, hilarious, and iconic moments,” while others omit disclaimers entirely.
Storyline Patterns
The storylines in these videos follow predictable patterns, often pitting a well - loved, usually older male celebrity (like Clint Eastwood, Denzel Washington, or Keanu Reeves) as the hero against a left - leaning talk - show host as the villain. This narrative plays on age - old tropes, appealing to a right - leaning, older audience. Videos targeting political figures also form their own sub - genre, indicating an underlying political element.
Denzel Washington recently expressed concerns about AI, stating that people are being misled by AI - generated content mimicking his voice, even though he has no social media accounts.
YouTube's Policies and Response
Policy Updates
On July 15, YouTube updated its policies to crack down on generative AI - made content. Content eligible for monetization must be “authentic” and “original,” though the policy doesn't explicitly mention generative AI. It only states that content must be the creator's original work and not mass - produced or repetitive.
Another policy requires creators to disclose when content makes a real person appear to say or do something they didn't, alters footage of a real event, or generates a realistic - looking scene that didn't occur.
Channel Removal
WIRED reached out to YouTube regarding over 100 AI - generated celebrity fanfic channels and policy enforcement. YouTube's Zayna Aston stated that all content must comply with Community Guidelines, and channels using deceptive practices, including misleading metadata, titles, and thumbnails, are not permitted.
WIRED can confirm that 37 flagged celebrity talk - show and fan - fiction - style channels, mainly those without AI disclaimers and some with egregious names like Celebrity Central and United News, were removed.
Creator Motives and Impact
Creator Rationales
One creator, speaking anonymously, claims to be creating fictional story interviews, combining storytelling, creativity, and a touch of realism. However, many channels, mostly based outside the US, likely have financial motives. Channels with names like “earningmafia” in their email and the presence of duplicated videos and multiple channels operated by the same creators suggest a profit - driven approach.
Platform - Driven Toxicity
Sandra Wachter, a professor in data ethics, AI, and algorithms at the University of Oxford, explains that rage - bait - style content is central to YouTube's business model. AI enables the cheap and rapid creation of such content, which keeps viewers engaged. While YouTube's move toward demonetization may have a positive impact, Wachter believes it doesn't address the root problem of a system that thrives on generating clicks.
Target Audience and Content Shifts
Most channels are based outside the US but target American celebrities, seemingly aiming to attract a trigger - happy and potentially lucrative online audience. Many channels have shifted focus from educational content (like cars, agriculture, or fitness) to AI - generated celebrity fan - fiction, possibly to game the algorithm or follow the latest trend for increased income.
Conclusion
These AI - generated “cheapfakes” on YouTube mark a transition. They don't rely on sophisticated special effects but rather on age - old psychological techniques and the platform's unique environment to deceive viewers. Their modus operandi seems capable of outsmarting YouTube's authenticity policies, as they aim to rage - bait, instigate debate, and trigger moral outrage. As long as the system rewards such content, it's likely that similar X posts will resurface, continuing to spread misinformation and toxicity. Manisha Krishnan contributed to the reporting of this story.
