DEV Community

robot1996
robot1996

Posted on

I built a first-pass AI video screening tool for suspicious social links

I built a small web tool for a problem I keep seeing: people encounter a suspicious social video and want a quick way to decide whether it deserves more verification before they share it.

The tool is called AI Video Detector. It accepts public social video links and uploads, then generates a first-pass evidence report.

The report focuses on cautious signals rather than final claims:

  • source/media access quality
  • visual artifact signals
  • motion consistency signals
  • compression signals
  • audio-video consistency where available
  • likelihood and confidence with limitations

The important product decision: the tool does not say it can prove whether a video is real or fake. It gives a structured signal report so users can decide whether to verify further, find the original source, or avoid sharing too quickly.

I’m looking for feedback on two things:

  1. Is the report language cautious enough for a sensitive category?
  2. Which signals would make the report more useful without overclaiming accuracy?

Top comments (0)