AI image detector
Upload an image and see whether it looks AI-generated — with the visual signals behind the call. An assessment, not proof.
Your image is not stored.
What it does
A vision model looks at the image and reports what it can actually see: lighting and shadows, skin and hair texture, hands, text, edges and reflections. You get a verdict, a confidence and the reasons.
Read this before you trust it
AI-image detection is not reliable. A single image can fool any detector, and a real photo can look synthetic after compression or editing. Treat the result as one signal among several — never as proof that an image is fake or real.
How it works
- Choose a JPEG, PNG or WebP image (up to 12 MB).
- The vision model inspects the pixels and lists concrete observations.
- You get a verdict, a confidence score and the reasons.
Questions
How accurate is it?
It is a judgement, not a measurement. Accuracy varies with the image, the compression and the generator. Use it as a hint, not a decision.
What do the verdicts mean?
"Likely AI" means the model found artefacts it associates with generation. "Likely real" means it found camera-like evidence. "Uncertain" means the evidence is mixed or missing.
Do you store my image?
No. The image is sent to the vision model for this one check and is not stored.
Can it identify who is in the photo?
No. The model is instructed not to identify people or infer private attributes.