Sri Lanka
USA
Canada
India
UK
UAE
TikTok has rolled back parts of its experimental AI feature after it began producing bizarre and sometimes incorrect summaries of videos, including cases where celebrities were misidentified as food items or objects.
The feature, known as “AI overviews,” was designed to appear beneath videos and provide automated descriptions or added context about what users were watching. It was initially tested with limited users in the United States and the Philippines.
However, the tool quickly drew attention after users began sharing screenshots of highly inaccurate and often absurd summaries involving popular creators and celebrities, including platform star Charli D’Amelio.
In one widely circulated example, a video of the dancer was described by the AI as a “collection of various blueberries with different toppings,” highlighting the system’s inconsistency. Other clips involving artists such as Shakira and Olivia Rodrigo were also reportedly misinterpreted with vague or incorrect descriptions.
The feature functioned in a similar way to AI-generated summaries seen in search engines, attempting to break down video content for users who expanded captions.
Following widespread criticism and viral posts on social media, TikTok has now limited the tool’s use, restricting it to displaying information about products featured in videos rather than general content descriptions.
The changes were first reported by tech publication Business Insider.
Some users also shared extreme misinterpretations of harmless content, including dance performances and comedy skits being described in violent or nonsensical ways. These errors led to online backlash, with many questioning whether the feature was experimental or unintentionally misleading.
The issues highlight ongoing challenges in generative AI systems, which can sometimes produce inaccurate or fabricated outputs despite rapid advancements in the technology.
Similar problems have been seen across the tech industry, including AI summary tools from major platforms generating misleading or humorous errors, prompting companies to adjust or temporarily suspend such features.
TikTok said it has identified the cause of the inconsistencies but did not provide full technical details, noting that user feedback played a role in refining the system.
