AI image detector methodology

How we check images

The tool combines a visual detector with a few basic file checks, then gives you a restrained result instead of a pretend certainty score.

What we look at

A standard scan uses a third-party visual AI-generation signal. It also reads the image format and dimensions, looks for readable metadata and provenance hints, and notes quality issues that could make the result shaky.

  • Visual AI-generation signal
  • Image format and dimensions
  • Readable metadata or provenance hints
  • Compression, low resolution, and screenshot warnings

How to read the three results

Likely AI means the available signals lean toward AI generation. Likely Real means they lean away from it. Uncertain means the file, the signals, or both do not support a reliable call.

What this tool does not do

It does not identify people, settle copyright, trace an image to its source, verify a news claim, or make a legal finding. Missing metadata does not count as evidence of AI generation.

Questions

Do you keep uploaded images?+

No. A standard upload is processed for the request and the original file is not kept.

What detector does the standard scan use?+

The current standard scan uses Sightengine for the visual AI-generation signal, then adds local quality and metadata checks.

Check an image