How our AI video check works
- You pick a video (MP4, MOV or WebM, up to 3 minutes).
- Your browser takes three still frames from the beginning, middle and end — the full video never leaves your device.
- We also read the start and end of the file for Content Credentials. OpenAI’s Sora and many other tools label their videos this way.
- Each frame runs through our AI image detectors and the results are combined into one verdict.
Signs a video is AI-generated
- Physics that doesn’t add up — objects passing through each other, liquids behaving strangely, footsteps that don’t match.
- Morphing details — faces, hands, text or logos that subtly change from frame to frame.
- Short, perfect clips — most AI videos are 5–20 seconds with smooth, cinematic camera moves.
- Watermarks — look for Sora, Veo, Kling, Hailuo or Runway logos, sometimes cropped or blurred out.
- No sound or generic sound — older AI clips are silent; newer ones may have audio that doesn’t quite sync.
Limits to know about
Video is harder than still images: frames are compressed, and AI video models are improving fast. Re-uploaded clips from TikTok, Instagram or X usually lose their labels. We cap confidence for videos at “High” unless a label confirms the origin, and we always show you the score for each frame.
Frequently asked questions
Can you detect Sora videos?
Yes. Downloaded Sora videos carry Content Credentials, which we read and confirm. Re-uploaded copies are checked frame by frame. See our Sora video detector page for details.
Do you upload my whole video?
No. Only three small still frames and a tiny slice of the file’s metadata are sent. The video itself stays on your device.
What video formats work?
MP4, MOV (iPhone) and WebM, up to 3 minutes. If your browser can play it, we can usually check it.
Is IsItMadeByAI really free?
Yes. Every scan is free, there is no sign-up, no credit card and no watermark on your results. The site runs on open-source detection models, which is how we keep it free.
How accurate is it?
When a file still carries Content Credentials or other AI labels (common for ChatGPT, Gemini, Firefly and Stable Diffusion files) the answer is essentially certain. Otherwise we combine several open-source detectors; they are right most of the time on unedited AI images but can be fooled by screenshots, heavy compression, filters and brand-new generators. That is why every result shows a confidence level and the evidence behind it — treat it as strong evidence, not proof.