Skip to content

A short tour

How a scan actually works

Think of this site as a bright little studio with one party trick: it looks at work and tells you, in plain English, how synthetic it seems.

Step one

You bring the file

No lab coat. No ten-step wizard. Drag a file onto the drop zone or paste a paragraph. Pictures, voice notes, PDFs, essays, even a frame from a clip. Scanning happens in this tab. The file is read to produce a score, then discarded. We only keep an anonymous filename, file type, score, verdict and time for site stats.

Step two

We look for tells

Text is read for cadence, hedging, and that too-balanced triad of adjectives. Pictures are checked for lighting, edges, texture, and the uncanny smoothness of a generated face. Audio is drawn as a waveform and, when we can, transcribed. Video is one still, not the whole clip.

Step three

The meter settles

You get a percentage — AI likelihood — a needle that lands on human, mixed, or AI, and a reason in ordinary English. Signals sit underneath like footnotes. It is a forensic estimate, not a court stamp. If we are unsure, we say mixed and we mean it.

What we will not pretend

Detectors can be wrong. Heavily edited humans fool us. Careful AI can hide. Use the score as a second opinion — for newsrooms, classrooms, hiring, and curiosity — not as a verdict to ruin someone’s week. Read the limits.

Try a scan