Free AI Image Detector
AI Image Detector checks pictures for evidence of AI generation without uploading them. It validates C2PA Content Credentials offline, reads IPTC digital source types and the prompts and settings saved by Stable Diffusion, ComfyUI, Midjourney and other generators, adds forensic hints and, if you want, an AI classifier score. Batch results export to JSON or CSV.
- Evidence:
- C2PA, AI metadata, hints, optional classifier
- Formats:
- JPG, PNG, WebP, AVIF, HEIC and more
- Batch:
- Up to 200 images, JSON or CSV report
- Price:
- Free
- Runs:
- Locally, in your browser
- Uploads:
- None: files never leave your device
- Account:
- Not needed
Quick answers
Can this tool prove that an image is real?
No tool can. It reports what the file says about its origin: signed Content Credentials, AI generation metadata and some hints. When nothing is found, the image may still be AI-generated, because social networks and messaging apps usually strip this data.
What are Content Credentials?
A C2PA signed record embedded by apps such as Adobe Firefly, Photoshop, ChatGPT, Microsoft Designer, Google and some phone cameras. It says which app made or edited the picture and whether AI was used. The tool checks the signature and the image hashes, and whether the signer is on the official C2PA trust list.
Processed on your device: your files are never uploaded. How it works
Network activity in this tab
What this page has requested since you opened the site in this tab. An upload is any request that sends data: POST, PUT, PATCH or a request with a body.
No uploads: nothing has been sent from this page.
No network requests yet.
To see every request, open your browser's developer tools (Network tab).
How to use the AI Image Detector
- Drop one or many images (JPG, PNG, WebP, AVIF, HEIC…) or paste one.
- Read the verdict of each image: Content Credentials, AI metadata or no provenance data.
- Open Details to see the signed history, the prompt and settings, the metadata and the file structure.
- Optionally download the AI classifier and press Classify for a statistical second opinion.
- Export the results of the whole batch as a JSON report or a CSV table.
Credits & open source
Our own code plus open-source libraries. Thank you to the people who make these projects.
- c2pa-web (Content Credentials)Reads and validates C2PA Content Credentials in images (WebAssembly, self-hosted from /vendor/c2pa). © Adobe and the Content Authenticity Initiative; built on c2pa-rs (MIT OR Apache-2.0) · MIT
- exifrFast EXIF, GPS, IPTC and XMP metadata reader. © Mike Kovařík, Mutiny.cz · MIT
- fflateFast ZIP compression for batch downloads. © Arjun Barrett · MIT
- ONNX Runtime WebExecutes ONNX models with WebGPU or WebAssembly (self-hosted from /vendor/ort). © Microsoft Corporation · MIT
Web platform APIs used: Web Workers, WebAssembly, WebGPU, File API, createImageBitmap, Canvas 2D, Clipboard API.
- Community Forensics DeepfakeDet ViT-Small (ONNX) by Jeongsoo Park and Andrew Owens, University of Michigan (weights); Han Yoon, Borderless / Ethix R&D (Hugging Face port and ONNX exports) (MIT) — optional AI-image classifier (8-bit 23 MB or full-precision 87 MB)
- C2PA trust list by Coalition for Content Provenance and Authenticity (C2PA) (CC-BY-4.0) — certificate authorities used to validate Content Credentials signers offline
Full license texts: third-party notices · all credits
Frequently asked questions
Are my images uploaded?
No. Everything runs in your browser. Links to credentials stored online are shown but not opened, and the optional classifier model is downloaded from Hugging Face only when you ask for it.
How reliable is the AI classifier?
It is a weak signal. It was trained on images from thousands of generators and works well on unedited outputs, but it makes mistakes on heavily edited, filtered, compressed or very new images. Never use it alone to judge a picture or a person.
Which generators are recognized from metadata?
Stable Diffusion tools (Automatic1111, Forge, ComfyUI, InvokeAI, Fooocus, SwarmUI), NovelAI and Midjourney descriptions, plus IPTC digital source types and generator names written by Adobe Firefly, OpenAI, Google, Microsoft, Meta and others.
Why does it say the credentials are broken?
The picture or the credentials changed after signing, so the recorded hashes no longer match. That often happens when an image is edited with software that does not update Content Credentials, but it can also be deliberate.