What it actually measures
A detector looks for statistical traces that generators and editors tend to leave: artifacts in the way pixels are arranged, inconsistencies in lighting or texture, patterns typical of GAN or diffusion models, or signs of splicing. From those signals it produces a probability, usually with a confidence level.
Danaya offers a free detector along these lines. You can drop in any photo, video or audio file and get a result, with one free check a day and no account needed. It is a screening tool: it tells you how suspicious a file looks, not a verdict carved in stone.
Where it falls short
A detector analyses a finished file, so it is always playing catch-up with the generators. As models improve, the traces they leave shrink, and detection gets less reliable. A clean-looking fake can slip through, and a genuine but heavily compressed or edited photo can be flagged by mistake.
The deeper limit is that a probability is not proof. It can tell you a file looks likely fake, but it cannot tell you a file is genuine, nor when or where it was taken.
Detection versus proof at the source
There is a second, stronger approach: prove a media is real at the moment it is captured, instead of guessing after the fact. Danaya seals a photo or video the instant it is taken, with a SHA-256 fingerprint of the file, the GPS location and a server timestamp, signed together. Capture is live only, so an imported or AI-generated file cannot earn a valid seal.
Anyone can then verify the seal from a public link. Where detection returns a probability, a seal returns a definite cryptographic match: the media is exactly what was captured, when and where it says. Use the detector to screen media you receive, and sealing to prove media you produce.