According to Business Insider, TikTok has recently scaled back testing of its "AI Overview" feature due to frequent severe hallucinations in generating video captions that resulted in multiple absurd gaffes. The function was originally designed to provide additional context for videos, recommend on-screen items, and explain visual content through text descriptions. While some videos received reasonable summaries, many cases proved disastrous. TikTok described a viral clip of popular creator Charli D'Amelio speaking alone against a white wall as an "assortment of blueberries with various toppings"; summarized a pet influencer's explanation about why dogs kick their legs after defecating as "intricately folded origami art on paper"; labeled Shakira promoting her new single as "multiple blue squares moving across the screen"; and characterized Olivia Rodrigo announcing her upcoming appearance on Saturday Night Live as "a person's face being gradually replaced by random letters and numbers." One Reddit user remarked that the feature seemed like it opened a separate tab after watching the video and generated accompanying text using a random text generator. TikTok confirmed to Business Insider that, following user feedback, the feature has been adjusted to shift focus from describing entire videos to identifying products within them. Previously rolled out to select users over several months in limited markets including the United States, TikTok declined to disclose whether it uses proprietary AI models or third-party solutions. This phenomenon of AI confidently fabricating nonsense is hardly unfamiliar—when Google launched its own AI Overviews in 2024, it similarly claimed with confidence that a dog had entered the NHL league. Yet witnessing beloved end-of-the-world-predicted AI technology stumble so hilariously offers a somewhat surprising yet comforting reassurance.