Visual analysis
A deep learning model examines the image as a whole: content, lighting, textures and details that generators struggle to render consistently.
How it works
Image generators leave subtle traces that the eye cannot see. VerifyGenAI combines two complementary analyses to estimate whether an image was created by AI.
The technology
Developed by Docusan, the model looks at an image the way an expert would: what it shows, and how it was made.
A deep learning model examines the image as a whole: content, lighting, textures and details that generators struggle to render consistently.
In parallel, the image is turned into forensic maps that reveal signals invisible to the eye, such as noise patterns and frequency signatures typical of generation.
Both views are merged into a single probability, from 0 to 100%, that the image was generated by AI.
From upload to verdict
Your image is sent over an encrypted HTTPS connection.
It is analyzed in memory on servers located in the European Union.
You get an AI-likelihood score and a verdict in seconds.
The image is not stored once the analysis is complete.
Reading the result
The score estimates the probability that the image was generated by AI. Here is how to use each verdict.
No action needed, unless other elements raise doubts.
The image is probably genuine. Stay attentive to the context.
Look for other evidence: source, metadata, original file, other pictures of the scene.
Treat the image as suspicious and check its origin before relying on it.
Do not rely on this image as genuine evidence without further checks.
Best results
Screenshots and copies downloaded from social networks are recompressed, which erases part of the traces.
Resizing and heavy compression reduce the signal available to the analysis.
Thumbnails and very small images contain fewer details to analyze.
Limits
No AI-image detector is infallible. New generators appear regularly and some images can be misclassified, in both directions.
VerifyGenAI estimates whether an image as a whole was generated by AI. It does not locate partial edits such as a pasted object or a retouched area.
For decisions with consequences, such as an insurance claim, an onboarding or an investigation, combine the score with other checks or rely on the complete forensic analysis of Docusan.
Discover DocusanTry it now
Upload an image or try our sample, and get a verdict in seconds.