Not everything can be taken at face value during the era of generative artificial intelligence. With AI video apps becoming more sophisticated, the internet is overflowing with hyper-realistic AI videos that can be indistinguishable from reality. Luckily, AI experts say there are a few ways you can help determine whether what you are looking at is real — or an extremely convincing fake.
Check for watermarks
One of the easiest ways to spot AI-generated videos is by watermarks. Videos made with Sora, OpenAI’s video generator, “include an easy-to-spot watermark, usually at the bottom left,” said PC Mag. Unfortunately, not all AI video apps include watermarks, and there are multiple ways to remove them, including cropping them out of the videos. In that case, it is crucial to look closer. Some removal tools are “nearly perfect or imperceptible, especially if the video is very simple,” Jeremy Carrasco, founder of Showtools.ai, said to Axios. Look for the “spongy block” where the watermark was removed.
Listen for garbled speech
AI experts say there are “telltale signs” of how the “voices and sounds in an AI video can often reveal its synthetic origin,” said HuffPost. The natural rhythm of real speech means some words are said slower than others, but “AI voices often sound unnaturally rushed all the time.” As people work out ways to spot AI-generated content, the em dash has become synonymous with ChatGPT-generated text. When asked about the equivalent in video, Bill Peebles, the head of Sora, said it was “this slightly wired speech pattern in Sora where it likes to say a lot of words quickly,” during an interview with video streaming show TBPN.
Because AI-generated speech has yet to master natural-sounding speaking rhythms, the voices generated by the apps often make “garbled sounds that appear to flatten out natural sound pitches,” said HuffPost. Human beings would never “produce that same kind of garbled quality, because, literally, we can’t,” Melissa Baese-Berk, a linguistics professor at the University of Chicago, said to HuffPost. Our vocal track can not “go from one sound to another” without some “blurring of the information between those two sounds.”
Check the metadata
It may seem tedious, but checking a video’s metadata will reveal its origins, and it is “easier to do than you think,” said CNET. Metadata is automatically attributed to content when it is created and can include “type of camera used to take a photo, the location, date and time a video was captured and the filename.” Every photo and video online has metadata, “no matter whether it was human- or AI-created.” Many AI-generated videos will also have “content credentials that denote its AI origins.”
Consider the content’s plausibility and source
One of the easiest ways to detect AI slop is to ask whether what you are seeing is even possible, Princeton University computer science professor Zhuang Liu said to Rolling Stone. If it is “not plausible in the real world, then it’s obviously AI-generated,” For example, “a horse on the moon or a chair made of avocado.” The impossibility means “these are obviously AI-generated,” he said. “That’s the easiest case.”
Next, check the source where you found the image. This does not “necessarily work for viral content,” especially since “they often come from previously unknown accounts,” but “seeing a video on a meme page could be a clue it’s not real,” said Rolling Stone.
Remain vigilant
Unfortunately, there is “no one foolproof method to accurately tell from a single glance if a video is real or AI,” CNET said. The best way to “prevent yourself from being duped” is to “not automatically, unquestioningly believe everything you see online.” Trust your gut instinct. If an item “feels unreal, it probably is.” In these “unprecedented, AI-slop-filled times,” your best bet is to “inspect the videos you’re watching more closely.”
Advanced AI may have made slop videos harder to spot, but experts say it’s still possible to detect them