For now, there are tell-tale signs that a video is AI-generated, rather than having been recorded in real life.
CLEVELAND — Sometimes it’s easy to spot a fake image or video and sometimes it’s not, so today, we’re VERIFYING how you can tell if a video was made using artificial intelligence — that is, at least for now.
We’ll use a video made by professional editor Yonatan Dor of the The Dor Brothers as an example, which has been viewed 1.1 million times and counting. Yonatan made it with the help of Grok, which is the AI chatbot developed by Elon Musk’s company, xAI.
The video makes it look like former President Donald Trump, Vice President Kamala Harris, Meta CEO Mark Zuckerburg, and others are committing crimes and being arrested.
We know these specific things didn’t happen, but if you weren’t sure, there are some tells in this video that would give it away. Yonatan told me over email, “With AI-generated videos, some common tells are strange-looking limbs, unrealistic movements, or missing/illogical details in the background. (T)hese subtle inconsistencies can give it away, but they will become harder to spot as the technology improves.”
When we look closely as his video, we see an example of strange looking limbs at around the 23-second mark, when the generated representation of Zuckerburg appears to be holding a bat. Around 30 seconds in, we also see a strange blur over what’s supposed to represent a police officer in the background.
But as Yonatan said, as technology improves, these tell-tale signs probably won’t show up any more. I asked him what he thinks we should do to prepare for this inevitability, and he told me:
“We believe there should be a law requiring every AI-generated video to have an invisible watermark. (T)his means that if you upload it to platforms like YouTube or Meta, the video would look completely normal — without any forced visible watermark — but the description would automatically include a label in an unremovable font stating that it’s AI-generated. This would ensure transparency without disrupting the viewing experience.
“We also think there should be a global committee similar to NATO for example, where nations can come together to agree on safety measures and regulations for AI development.”
For now, the best advice we can give you is pay attention to the details, and if you see something you’re not sure about, to send it to us at VERIFY@wkyc.com.
