1. Look for Physical and Facial Inconsistencies
Even with advanced technology, AI often struggles with the subtle complexities of the human body and the physical world. Pay close attention to the details. Hands and fingers are a classic giveaway; AI can render them with unnatural proportions or an incorrect
number of digits. Similarly, look at faces for unnatural movement. AI-generated people may blink infrequently or at a strange rhythm, and their eyes can sometimes appear glassy or have a fixed stare. Other details like messy hair, teeth, and ears can appear blurry, overly smooth, or asymmetrical. Pause the video and examine still frames for these minor but revealing errors.
2. Listen for Audio Anomalies and Bad Lip-Sync
Your ears can be just as crucial as your eyes for detecting fakes. People often forgive imperfect visuals, but bad audio is an immediate red flag. Listen for a robotic or flat tone, stilted pacing, or a lack of emotional expression in the voice. Another major sign is poor lip-synchronisation, where the mouth movements do not quite match the spoken words. This mismatch often becomes more pronounced as a sentence progresses. The audio might also contain strange artifacts like static, warbles, or inconsistent background noise, which suggests the sound was generated or manipulated separately from the visuals.
3. Scrutinise the Background and Lighting
What is happening behind the main subject is often as important as the subject itself. AI generators can struggle with maintaining consistency in the background. Look for objects that seem to warp, shimmer, or blur, especially during movement. Text in the background, such as on signs or packaging, is another weak point. AI often renders text as indecipherable or morphing strings of characters. Pay attention to the lighting and shadows. Do the shadows fall correctly based on the light sources in the scene? Inconsistent lighting, where a person is lit differently from their environment, can expose a composite video.
4. Check the Source and the Context
Before you even analyse the pixels, investigate the source of the video. This is often the fastest way to debunk a fake. Ask yourself: who is sharing this video, and why? Is the account known for posting reliable information, or is it a new or anonymous account? Look at the account's history for other posts. A credible news event will almost always be reported by multiple trusted sources, not just one random account on social media. If a video is designed to provoke immediate outrage or a strong emotional reaction, it is a good reason to be extra skeptical and double-check its authenticity before believing or sharing it.
5. Use Reverse Image Search
While there is not yet a simple reverse search tool for entire video files that is widely and freely available, you can do the next best thing by searching for keyframes. Take screenshots of the most important moments in the video and use a reverse image search tool like Google Images or TinEye. This can help you determine if the video, or parts of it, have appeared online before. Often, a viral video is not a sophisticated AI creation but simply an old, real video that has been repurposed with a false and misleading context. A reverse image search can quickly uncover the original source, revealing where and when it was actually filmed and for what purpose.
