How to Tell if a Video Is AI-Generated (Sora, Veo, Kling)
AI video tools like OpenAI Sora, Google Veo, Kling and Runway now produce clips that fool most viewers. Watermarks help, but removal tools are everywhere. Here's how to check a suspicious video properly.
1. Scan the file for Content Credentials
Sora videos carry C2PA Content Credentials naming OpenAI. Download the original MP4 and drop it into our Sora video detector. Re-encoding by social platforms often strips this data, so the original file matters.
2. Check for SynthID
Google Veo videos carry an invisible SynthID watermark. Google's free SynthID Detector can verify it and shows which segments are marked. See what SynthID catches and misses.
3. Don't trust a missing watermark
Visible watermarks can be cropped or painted out in seconds. Their absence tells you nothing on its own.
4. Look for frame-to-frame glitches
- Objects that change shape, appear or vanish between frames
- Text on signs or clothing that shifts from shot to shot
- Hands, teeth and faces that smear during fast motion
- Physics that is slightly wrong: liquid, hair, fabric, collisions
- Short clips (often 5–20 seconds) with no wider footage of the same event
5. Trace the source
Find the earliest upload. Real news footage usually appears from several independent angles and accounts. A single viral clip from a new account deserves suspicion.