1. Look for Physical Imperfections
Artificial intelligence models are powerful, but they still struggle with the complex and subtle details of the physical world. The first place to look for giveaways is in the anatomy of people on screen. Hands are a classic failure point; you might see
extra fingers, unnatural bends, or hands that seem to flicker or warp during gestures. Hair can also be a tell, sometimes moving as a single, solid block rather than individual strands. Other times, you may notice that jewellery or glasses have an unnatural appearance or that reflections don't behave as they should in the real world. Skin can appear overly smooth and waxy, lacking the natural pores and textures visible in high-resolution footage. While these errors are becoming less common as technology improves, they are often the first and most obvious sign that what you're watching wasn't filmed with a camera.
2. Check the Eyes and Face
The face is often the focus of an AI video, but it's also a source of key inconsistencies. Pay close attention to the eyes. Unnatural blinking patterns—too fast, too slow, or completely absent—have long been a sign of a deepfake. Though some newer models have improved this, many still get it wrong. Also, look for a strange or flickering blur around the edges of the face, especially where it meets the hair or neck, which can indicate a face has been digitally placed onto a body. Finally, watch for facial expressions that don't quite match the emotional tone of what's being said. The AI might generate a technically perfect face, but it often fails to capture the subtle muscular movements that create a genuine human expression, resulting in a slightly 'off' or emotionally flat appearance.
3. Listen for Audio and Sync Issues
Your ears can be just as important as your eyes when detecting AI content. Synthetic voices can sound robotic, monotone, or have an unnatural cadence and rhythm. The audio might also be suspiciously clean, lacking the normal background noise you'd expect in a real-world recording. Another major red flag is poor lip-syncing. Watch the person's mouth closely to see if the words align perfectly with their lip movements. In many AI-generated videos, particularly those made quickly, there is a noticeable lag or mismatch, which becomes more obvious during rapid speech. While lip-sync technology is advancing, any slight desynchronization should be a cause for suspicion.
4. Investigate the Source and Context
Beyond visual and audio cues, one of the most powerful verification methods is to investigate where the video came from. This is a critical step in digital literacy. Ask yourself: Who posted this content? Is it a reputable news organization, a known public figure, or an anonymous account with no history? Check the account's profile, its past posts, and whether it has a history of posting sensational or unverified claims. Some platforms, like YouTube, have labels for altered or synthetic content, so be sure to look for those disclosures. You can also perform a reverse image or video search using tools like the InVID/WeVerify browser extension to see if the clip has appeared elsewhere online in a different context. Often, misinformation is old footage that has been re-shared with a new, false narrative.
5. Pause and Check Your Emotions
Misinformation is often designed to provoke a strong and immediate emotional reaction, like outrage, fear, or shock. These feelings can override our critical thinking and push us to share content before we’ve properly vetted it. If a video makes you feel a powerful urge to hit 'share' immediately, that's a signal to pause. This emotional response is a tactic. By creating a sense of urgency, creators of fake content hope you'll spread their message without thinking. Before you amplify a clip, take a moment to run through the other checks. Ask why the video was made and who benefits from its spread. This simple habit of pausing to question your own reaction is a crucial defense against becoming an unintentional purveyor of fake news in an increasingly complex digital world.
