Drop in an image and check it for AI traces: embedded Content Credentials (C2PA), generator signatures and camera metadata. An honest result — including a clear note when no definitive traces can be found.
100% in your browser · your image is never uploaded
Reads embedded C2PA data and generator signatures (DALL·E, Midjourney, Stable Diffusion, Firefly …). When present, the origin is definitive.
Real photos usually carry EXIF data: camera model, aperture, shutter speed, ISO, GPS. Such traces point to a genuine photograph.
Generator-typical resolutions and missing metadata feed in as weak hints – clearly flagged as such.
When no metadata is left, a careful look helps. These are the most common giveaways of AI images:
Too many or fused fingers, unnatural joints – the classic weakness of image generators.
Warped, invented or unreadable letters on signs, packaging or clothing.
Asymmetric pupils, too many teeth, oddly shaped ears or jewellery.
Merging objects, impossible architecture, patterns that repeat senselessly.
Shadows that don't match the light source, or reflections that don't add up.
Overly smooth, waxy skin and a plastic-like sheen with no real pores or detail.
No detector is perfect. It's only definitive when the image carries provenance data (Content Credentials) or generator signatures. Without those, AI can only be estimated from weak hints – modern generators are often photorealistic.
No. The entire analysis runs in your browser. Your image never leaves your device and is not stored.
An industry standard for provenance: many AI tools and cameras (OpenAI, Adobe, Google, Leica …) embed signed data showing how an image was made. We read this data when it's present.
Social networks and messengers strip almost all metadata on upload. Without those traces only a visual estimate remains – and we deliberately flag that as uncertain.
We create sales-ready product images, listings and campaign visuals with AI – clean, on-brand and at scale.