By Harshit, LONDON, November 5, 2025
Social media feeds are increasingly dominated by AI-generated videos, and distinguishing between real footage and artificial content is becoming a major challenge for users worldwide. Experts say that, for now, one of the clearest signs of AI content is deceptively simple: poor video quality.
In the past six months, AI video generators have improved dramatically, creating content that can mimic human filming with alarming accuracy. Yet subtle visual cues remain that can help viewers identify fakes — at least temporarily.
“It’s one of the first things we look at,” said Hany Farid, computer-science professor at the University of California, Berkeley, and founder of the deepfake detection company GetReal Security. Farid explained that grainy, blurry, or pixelated footage is often an early warning sign that a video may have been AI-generated.
While low-quality footage is not definitive proof of AI manipulation, it increases the likelihood of spotting anomalies. Matthew Stamm, head of the Multimedia and Information Security Lab at Drexel University, stressed: “If you see something that’s low quality, it doesn’t automatically mean it’s fake — but it’s a clue to examine it more closely.”
The Subtle Artifacts of AI Videos
Advanced AI video generators, such as Google’s Veo and OpenAI’s Sora, have moved beyond obvious mistakes like extra fingers or garbled text. Instead, they often introduce subtle distortions: unnaturally smooth skin, irregular hair patterns, slight shifts in clothing, or background objects that move unrealistically. These inconsistencies are easier to detect in high-resolution footage.
Ironically, low-quality videos can mask these flaws, making AI-generated content more convincing. Many viral AI clips, from bunnies on trampolines to staged subway romances, intentionally use grainy, security-camera-style visuals to hide artefacts.
Farid outlined three factors that commonly indicate AI-generated videos: resolution, quality, and length. Short clips — typically six to ten seconds long — are the most prevalent because producing longer AI videos is expensive, and errors accumulate with duration. “You can stitch multiple AI clips together, but cuts every few seconds often give them away,” he added.
Low Resolution and Compression as Concealment
Video resolution refers to the pixel count, while compression reduces file size by discarding detail. AI creators sometimes deliberately downgrade these properties to hide imperfections.
“If I’m trying to fool people, I generate a fake video, reduce the resolution, and add compression,” Farid said. The technique is common in viral AI content, allowing viewers to enjoy the video while missing subtle cues that it’s synthetic.
However, as technology advances, even these visual tells may disappear. Stamm noted that within a few years, many AI-generated videos may be indistinguishable from real footage without advanced verification methods.
The Importance of Provenance
Experts stress that the long-term solution is digital literacy and verification, not simply spotting imperfections. AI videos, like text, require viewers to consider the source, context, and verification rather than trusting appearances.
Mike Caulfield, a digital literacy researcher, suggested: “Video is going to become somewhat like text. Provenance, not surface features, will be the most critical factor. People must learn to treat all visual media with scrutiny.”
Tools are emerging to help verify content. Statistical “fingerprints” left behind by AI generation can sometimes reveal manipulation, and companies are developing methods to embed origin information in both real and AI content.
A Growing Information Security Challenge
As AI-generated video becomes ubiquitous, experts warn it represents one of the greatest information security challenges of the 21st century. Stamm said the field is still young, with a relatively small but rapidly expanding number of researchers tackling the problem. He added: “We need a combination of education, policy, technology, and awareness to address it, but I’m not prepared to give up hope.”
For now, the safest approach for social media users is caution: assume videos can be fake, check the source, and verify context. In an age where even moving images can be manipulated seamlessly, critical thinking remains the most reliable defense.