Analyse the Visuals: Look for Glitches
The first step is to scrutinise what you are seeing, as AI models often leave behind subtle errors, or 'artifacts'. Pay close attention to the face. Eyes can be a major giveaway; look for unnatural blinking patterns, a lack of blinking, or eyes that don't
seem to focus correctly. In a real video, the reflections in both eyes should match, but deepfakes often fail to replicate this. Another area to watch is the edge of the face and hair, where you might see blurring, flickering, or an unnatural seam where the fake face is overlaid onto a real head. AI also struggles with rendering fine, repetitive details like individual teeth or strands of hair, which may appear blurry or unnaturally perfect. Finally, check for inconsistencies in lighting and shadows. If the light on the person's face doesn't match the lighting in the surrounding environment, it’s a significant red flag.
Listen for Audio Oddities
Deepfake creators often focus more on the video than the audio, which can be a weak point. Listen for a voice that sounds robotic, flat, or lacks emotional expression, even when the topic is emotional. The rhythm and pacing might feel unnatural, or you might notice a lack of breaths and other normal mouth sounds between words. Another key sign is poor lip-syncing, where the mouth movements don't perfectly align with the spoken words. Sometimes, the background noise can also be a clue. It might sound too clean, like it was recorded in a studio, or cut in and out abruptly, which is inconsistent with a real-world recording.
Pause and Question the Context
Beyond the technical flaws, one of your most powerful tools is simple critical thinking. Slow down and consider the content of the video. Does the message seem plausible? Is it something you would expect this person to say or do? Malicious deepfakes are often designed to provoke a strong emotional reaction—like outrage or excitement—to cloud your judgment. Ask yourself why this video is appearing now and what it's trying to make you do or believe. If a video of a public figure makes an outrageous claim, or a supposed relative urgently asks for money or sensitive information, your skepticism should be high. This contextual check is often more reliable than spotting visual flaws, especially as the technology improves.
Verify the Source, Not Just the Video
Verifying where a video came from is just as important as analysing the video itself. Before you share or believe a video, investigate its origin. Who posted it? Is it a reputable news organisation or a random, anonymous account? If a video claims to show a breaking news event, check to see if trusted news outlets are also reporting it. You can also use a reverse image search by taking screenshots of the video and uploading them to tools like Google Images or TinEye. This can help you see if the video has appeared online before, potentially in a completely different context. If the original source acknowledges the video is a parody or experiment, but it's being shared as real, you've found your answer.
