With the ongoing impeachment trial of Vice President Sara Duterte bringing renewed attention to the November 2024 Zoom conference video cited under Article IV, discussions have once again raised questions about whether videos can be trusted or dismissed as AI-generated. As artificial intelligence becomes increasingly capable of producing convincing fake footage, here’s what to look for before concluding that a video is real or AI-made.
Pay close attention to facial movements; watch how often they blink, whether their lip movements match the audio, and how their facial expressions change throughout the video. According to Sutherland, AI-generated faces often appear less emotionally expressive than real people. Real people blink 15-20 times a minute, with natural variation tied to what they’re thinking and focusing on. While AI faces tend to blink too rarely, too often, or at suspiciously regular intervals.
Some fake videos can also be caught with the small mismatches between mouth shapes and audio. When it comes to facial mannerisms, real faces have over 40 muscles working in slightly asymmetrical, imperfect ways; while AI often looks a little too smooth or too symmetrical, especially as a clip goes on.
Hands and the surrounding environment can also provide clues. Although AI has improved significantly, it may still struggle to generate realistic hand movements. Look for extra digits, fused fingers, or hands that move stiffly and don’t quite match what’s being said. Also watch for objects that appear, disappear, or morph into something else mid-frame, and edges where a person’s body blends unnaturally into the background.
In authentic footage, shadows, reflections, and light sources generally remain consistent throughout a scene. AI-generated videos may contain shadows falling in the wrong direction, reflections that do not match nearby objects, or faces that appear to be lit differently from the rest of the environment.
Some platforms such as Meta and Google also label AI-generated or significantly altered content. Videos created using certain AI tools may include watermarks or disclosures indicating they were generated or edited with artificial intelligence. However, not every AI video is labeled, so the absence of a disclosure should not be taken as proof that a video is authentic.
More than the visuals, listening to the audio and hearing voices that sound unusually flat, robotic, or disconnected from the speaker’s emotions also helps. Background sounds should match the setting.
Aside from the aforementioned characteristics, the most reliable way to verify a suspicious video is not by looking for visual flaws, but by checking the claim it makes. If a clip appears to show a public figure making a controversial statement or captures a supposedly breaking event, look for reports from credible news organizations and try to trace the footage back to its original source. Videos designed to provoke outrage, fear, or urgency should always be approached with extra caution.
As AI technology continues to improve, visual clues alone are becoming less dependable. The strongest defense against misinformation remains verifying the source, confirming the context, and thinking twice before sharing.
In regards to Vice President Sara Duterte’s case, whether a video is authentic or manipulated cannot be settled by social media speculation alone. Like any piece of evidence, its authenticity must be established through verification and the evidence presented before the Senate impeachment court.
