On TikTok, creators have stumbled upon a simple, reliable way to rack up millions of views—exposing the bizarre flaws of generative AI.

Hilarious conversations with ChatGPT, Grok and Claude are going viral, with TikTokers exposing how LLMs flatter and deceive users, even in response to simple queries.

AI Is Perfect For TikTok Comedy Skits

Billions of dollars have been invested into generative AI, but there’s a fatal, hilarious flaw at the heart of the technology which cannot be fixed—hallucinations.

Despite breathless predictions of incoming “AGI” by tech-enthusiasts, generative AI is not sentient—it’s not even close—LLMs simply do not know the difference between truth and nonsense.

Husk is one of the most well-known creators who has managed to monetize the errors of AI—his deadpan delivery makes the perfect comedy foil to the absurdity of LLMs.

Husk’s content mirrors the classic comedy dynamic of the “straight man” reacting to the confident delusions of a fool, with AI playing the part perfectly.

Watch enough of Husk’s videos, and AI begins to resemble a sitcom character, a dangerously incompetent, weird butler who has no idea what he’s talking about, who will shamelessly flatter and lie to Husk in order to save face.

On TikTok, AI insists that December is spelt with an X, gives a positive review of a non-existent “song,” and reacts to life-threatening situations with calm ineptitude.

One viral video (viewed more than nine million times on TikTok), sees Husk ask ChatGPT to time how long it takes to run a mile.

The chatbot dutifully responds, telling Husk that it has set a timer. When Husk instructs the AI to stop the timer after a few seconds, ChatGPT tells him that it has been ten minutes, and argues that its timing is accurate.

The “fake timer” video even made its way to OpenAI CEO Sam Altman, who explained the error as the inability of that particular model to set a timer.

In response, Husk filmed a follow-up video, making AI react to Altman’s reaction.

Amusingly, ChatGPT stubbornly insists that it does have the ability to set an accurate timer, despite Altman’s statement, then gives another woefully inaccurate measurement.

TikTok creator @fatherphi is also racking up millions of views with the same bit, cycling through different models to expose hilarious hallucinations.

One video sees ChatGPT give flattering, detailed feedback in response to a revolting audio clip of fart sounds, and suggests that the “song” could potentially do well on Spotify.

YouTube creator Jonas Čeika tried the same experiment, and got the same result.

Creatives (especially musicians) who are using LLMs for feedback should take note—asking a real person for their honest opinion is surely better than asking a dead, indifferent machine to rate a piece of art.

The TikTok trend highlights the public’s growing dissatisfaction with AI, and the fierce backlash against the boom of polluting, power-hungry data centers that are integral to the error-riddled technology.

An NBC poll found that a mere 26% of Americans had positive feelings about AI, with around half of the respondents having negative feelings—Gen Z was found to hold a deep distaste for the technology.

The negativity is hardly surprising, considering that AI was pitched as a technology powerful enough to automate jobs and displace workers, sparking fears of apocalyptic harm in the most hyperbolic predictions.

This framing even led to the adoption of the phrase, "permanent underclass,” referring to a hypothetical group of workers doomed to poverty by the technology.

Thanks to TikTok, the perceived threat of generative AI is becoming a punchline.

MORE FROM FORBES

ForbesSam Altman Is Starting To See The Dead Internet TheoryForbesWhat Does ‘You The Birthday’ Mean? TikTok’s Viral Phrase, ExplainedBy Dani Di PlacidoForbesDisney’s Animatronic Glitches Spark A Wave Of MemesBy Dani Di PlacidoForbesWhy Did JD Vance Compare Aliens To Demons?By Dani Di Placido