Upload any photo and check it for AI-generation signals — embedded generator metadata, content credentials and pixel-level forensics.
Drop a PNG, JPG or WebP (max 15 MB). Results are an honest estimate — never proof.
Drop an image here or click to browse
PNG · JPG · WebP — stays on your device
Bright areas were compressed differently from the rest — a classic sign of editing or AI inpainting. Uniform grey = natural.
Many AI tools stamp files with hidden metadata — Stable Diffusion parameters, DALL-E tags, Firefly markers. We read the raw file bytes to find them.
We look for C2PA/JUMBF manifests — the tamper-evident labels tools like Adobe Firefly embed in AI-made images.
Error Level Analysis recompresses the image and highlights regions with abnormal compression — where edits or AI fills often hide.
It's a strong first check, not a verdict. Metadata can be stripped and ELA can misread heavy filters. Always verify important images through the source.
No. All analysis happens in your browser with JavaScript — the image never leaves your device.
Google's SynthID check needs their official detector. We detect the signals a browser can read: metadata, C2PA and compression forensics.