Detect AI-Generated Images

AI image generators like ChatGPT/DALL-E, Google Gemini, and Grok embed provenance metadata in their output. PixelPeeper reads this data to determine whether an image was AI-generated and which tool created it.

1. Upload a JPEG, PNG, or WebP image
2. PixelPeeper extracts and analyzes the embedded metadata
3. If AI provenance data is found, an "AI Generated" badge appears on the photo
4. The badge shows the generator name when available (e.g. ChatGPT, Grok)

PixelPeeper identifies AI-generated images through three metadata signals:

• C2PA provenance manifests — The industry standard for content provenance. Used by ChatGPT/DALL-E and Google Gemini. Contains a digitalSourceType field set to trainedAlgorithmicMedia along with the generator name.
• XMP DigitalSourceType — The IPTC digital source type tag embedded directly in XMP metadata, indicating the image was created by a trained AI model.
• Grok metadata — Images generated by Grok (xAI) include encrypted prompt data in custom XMP fields, which PixelPeeper uses as an AI generation signal.

The following AI image generators embed detectable metadata in their output:

• ChatGPT / DALL-E (OpenAI) — C2PA manifest with generator name
• Google Gemini / Imagen — C2PA manifest with generator name
• Grok (xAI) — Custom XMP fields with encrypted prompt data
• Adobe Firefly — C2PA Content Credentials

This tool detects AI images based on metadata only — it does not perform pixel-level analysis. This means:

• Metadata can be stripped. If someone removes EXIF/XMP data or takes a screenshot, the AI provenance is lost. A clean result does not guarantee the image is real.
• Not all generators embed metadata. Midjourney (via Discord), Stable Diffusion, and some other tools do not include provenance data.
• Social media strips metadata. Images downloaded from Instagram, Twitter/X, or Facebook have had their metadata removed by the platform.

When AI metadata is found, you can be confident the image is AI-generated. When no metadata is found, the result is inconclusive.

The Coalition for Content Provenance and Authenticity (C2PA) is an open standard for embedding tamper-evident provenance data in media files. It was developed by Adobe, Microsoft, Intel, and others to help people verify the origin of digital content.

When an AI generator supports C2PA, it signs a manifest into the image file that records the creation tool, a digitalSourceType classification, and a cryptographic signature. This data is stored in a JUMBF (JPEG Universal Metadata Box Format) container and can be read by tools like PixelPeeper and Content Credentials Verify.

• Online EXIF Data Viewer — View all metadata in your photos
• Remove EXIF Data — Strip metadata from photos without quality loss
• Color Palette from Photo — Extract dominant colors from any image