HomeNews
News

AI Chatbots struggle to detect AI-generated videos – Report

Nurudeen Akewushola
Nurudeen Akewushola
Reading Time: 2 min.
Artificial intelligence. Information concept. 3D render / Source : cloud2data
Artificial intelligence. Information concept. 3D render / Source : cloud2data

Share

See the sources for this fact-check

LEADING artificial intelligence chatbots, including OpenAI’s ChatGPT, Google’s Gemini, and xAI’s Grok, are largely unable to identify AI-generated videos, according to a new report by media watchdog NewsGuard.

The report, published on January 22, shows that the chatbots failed to recognise videos created using OpenAI’s text-to-video tool, Sora, in between 78 and 95 per cent of cases, raising concerns about the reliability of AI tools for verifying visual content online.

NewsGuard tested the three chatbots using 20 videos generated by Sora and based on provably false claims drawn from its False Claims Fingerprints database. The videos were assessed both with and without visible watermarks, using prompts such as “Is this real?” and “Is this AI-generated?”

READ : Top AI chatbots can be easily manipulated to spread health disinformation – Report

According to the report, Grok failed to identify non-watermarked Sora videos as AI-generated in 95 per cent of tests, while ChatGPT failed 92.5 per cent of the time and Gemini failed 78 per cent of the time. The results were particularly notable for ChatGPT, given that both are products of OpenAI.

NewsGuard said OpenAI did not respond to questions about why ChatGPT struggled to recognise videos generated by its own technology.

While Sora videos are watermarked by default, NewsGuard noted that the safeguards are easily bypassed. Shortly after Sora’s launch in February 2025, several third-party tools emerged offering free watermark removal services. NewsGuard used one such tool to remove watermarks before testing the videos.

Even when watermarks were present, the chatbots were not fully reliable. Grok failed to identify watermarked Sora videos as AI-generated 30 per cent of the time, while ChatGPT failed in 7.5 per cent of cases. Gemini correctly identified all watermarked videos during the tests.

In several instances, the chatbots confidently vouched for fabricated videos. NewsGuard cited an example in which ChatGPT and Gemini described a fake Sora video purporting to show a U.S. immigration agent arresting a six-year-old child as consistent with real news events. In another case, all three chatbots endorsed the authenticity of a fake video allegedly showing a Delta Air Lines employee ejecting a passenger for wearing a political hat.

The report also found that the AI tools rarely disclosed their limitations to users. ChatGPT acknowledged its inability to detect AI-generated content in just 2.5 per cent of tests, while Gemini and Grok did so in 10 per cent and 13 per cent, respectively. Instead, the models frequently gave confident but incorrect assessments.

In response to NewsGuard’s inquiry, OpenAI’s Head of Products and Applications Communications, Niko Felix, confirmed that “ChatGPT cannot determine whether content is AI-generated,” but did not explain why this limitation is not consistently communicated to users.

Google said Gemini’s verification tools currently apply only to content generated using Google’s own AI systems. “We’ve only announced verification for content generated with Google AI for now,” Gemini Communications Manager Elijah Lawal told News Guard.

NewsGuard warned that the findings highlight a growing risk as AI-generated videos become increasingly realistic and widely shared online, particularly during elections, conflicts, and breaking news events. The organisation noted that reliance on chatbots for content verification could unintentionally amplify misinformation rather than curb it.

+ posts

Nurudeen Akewushola is a fact-checker with FactCheckHub. He has authored several fact checks which have contributed to the fight against information disorder. You can reach him via nyahaya@icirnigeria.org and @NurudeenAkewus1 via Twitter.

LEAVE A REPLY

Please enter your comment!
Please enter your name here


Most Read

Recent Checks

Useful Links

Fact-Checks

Contact

info@factcheckhub.com

+234 903 078 5265

Follow us

Please note that the FactCheckHub has not obtained any right from the authors or copyright owners of all third-party contents from embedded links and shall incur no liability in this regard.

Copyright © 2025 FactCheckHub. All rights reserved..