A Practical Guide to Spotting AI Fakes
See how you can track each cable, above buttons and letters, where it's going, being consistent in its form and color? Well, spotting AI fakes is a lot like this exercise. Check this out:
Not gonna lie, I spend a lot of time thinking about the tools that shape our reality. For generations, the camera has been a bastion of proof. "Seeing is believing," we'd say. But we are now firmly in an era where that is no longer true. The rapid rise of generative AI video means that realistic, entirely fabricated videos can be created with increasing ease. This isn't a distant, sci-fi future; it's happening right now, and it has profound implications for misinformation, propaganda, and even our personal security.
The erosion of trust in visual media is a serious challenge. How can we function as a society if we can't believe video of our leaders? How do we protect ourselves when a loved one could seemingly appear in a video asking for money in an emergency?
While the problem is complex, empowering ourselves with knowledge is the first and most critical step. The brilliant team at Corridor Crew recently released a video, "Can We Teach our Moms to Spot Fake Ai Videos?", that breaks down this challenge beautifully. Using their insights, I've put together this practical guide to help you develop the essential skill of digital discernment.
Your Checklist for Reality: 5 Things to Look For
Think of this as a mental checklist to run through whenever you encounter a video that feels even slightly "off". You can even save this article, so you can open this checklist for when you need it.
1. Check the Upload Date (The 2023 Rule)
This is the simplest and, for now, one of the most effective checks. High-quality generative AI video technology wasn't widely available before 2023. If you see a stunning, seamless video supposedly from 2021 or earlier, your suspicion should be high. It's a simple piece of context that can immediately debunk many fakes.
2. Count the Seconds
Creating AI video is computationally expensive. As a result, most AI-generated clips are incredibly short – often just 5 to 10 seconds long, and almost always consisting of a single, continuous shot without any cuts. If a video is several minutes long with multiple camera angles and edits, it's far more likely to be real. A short, impressive, single-shot clip, however, warrants a closer look.
3. Scrutinize the Fine Details (Text & Teeth)
AI models are incredible mimics, but they struggle with complex, consistent details that we take for granted. Two major giveaways are:
Text: Look at any text in the background – on signs, books, or screens. In an AI video, it will often be a garbled mess of nonsensical characters or misspelled words.
Teeth: This is a fascinating one. AI is notoriously bad at creating realistic teeth. They often appear blurry, oddly uniform, inconsistent from frame to frame, or just generally unnatural. It's a small detail that a human filmmaker would get right, but an AI model often fails.
4. Look for Breaks in Continuity
This is a huge tell. AI video is typically generated shot by shot, without an overarching understanding of the entire scene. This leads to major continuity errors:
Character Continuity: A person's appearance might subtly change. Their shirt color could shift, or their earrings might appear and disappear.
Environment Continuity: Look at the background. An object on a table in one shot might be gone in the next, or the layout of a room might not make logical sense from one angle to another.
Action Continuity: Actions may not flow logically. A person might be reaching for a cup, and in the very next shot, the cup is already in their hand with no connecting movement.
5. Question the Logic
This is where your own common sense is your best tool. AI doesn't understand the world or the laws of physics the way we do. Watch for things that are simply illogical. Do cars drift through each other? Do shadows fall in the wrong direction? Does an object appear from nowhere? As Corridor Crew notes, it's like an optical illusion where the lines don't actually connect in a way that makes sense in three-dimensional space.
The Ultimate Tool: Critical Thinking
These technical tells are incredibly useful, but they have a shelf life. As AI technology improves, it will get better at teeth, text, and continuity.
The single most important tool, which will never become obsolete, is critical thinking. The final, and most crucial, step in verification is to always seek a second source.
See a shocking video of a public figure? Don't share it. Open a new tab and check with multiple, reputable news organizations. If it's real, they will be reporting on it.
Get a panicked video message from a family member asking for money? Hang up and call them back on their known phone number. Verify it directly, or my personal favorite - call a tech-savvy person and ask them to take a look as well. Chances are they’ve seen many more such fakes, and are able to spot them faster, if you’re concerned that time is of importance in that situation.
Developing this habit of healthy skepticism it's a fundamental life skill for every citizen in the 21st century. By learning to spot these fakes, we not only protect ourselves from deception but also contribute to a more resilient and truth-based digital society.