AI Metadata Checker
Inspect C2PA manifests, XMP namespaces, EXIF parameters, and PNG workflow chunks associated with AI image exports in browser RAM.
Drop your image here to check AI metadata
Drag & drop a JPG, PNG, or WebP file (or up to 30 images for batch checking), or click to browse.
Deep Technical Metadata Inspection
Modern digital images contain multiple layered metadata dictionaries. This tool maps out individual field values across C2PA Content Credentials, XMP schemas, EXIF tags, and PNG text chunks.
Frequently Asked Questions
How does the AI Metadata Checker differ from the AI Image Checker?
The AI Image Checker provides a rapid provenance synthesis (summarizing whether C2PA, generator tools, or software signatures are detected). The AI Metadata Checker provides a detailed technical field-by-field breakdown of every supported tag, parameter string, and raw chunk for in-depth file analysis.
Can I view the raw generation prompt text here?
Yes. If the image was exported by a tool that embeds prompt strings (such as Automatic1111, Midjourney web, or ComfyUI), this checker will parse and present the full prompt and negative prompt text.
Does checking metadata alter the file in any way?
No. Inspection is strictly read-only. Your original image is read into volatile browser memory and remains unaltered unless you choose to execute a cleaning operation.
What specific AI metadata formats can this tool detect?
This tool detects C2PA Content Credentials (JUMBF manifests and X.509 certificates), Stable Diffusion text parameters (in PNG tEXt/iTXt chunks), ComfyUI execution graphs and workflow JSON, Midjourney web metadata, DALL-E/ChatGPT signatures, Adobe Firefly provenance assertions, and IPTC/XMP generative tags.
Can this tool detect whether an image is AI-generated if all metadata has been removed?
No. This tool operates strictly on embedded container metadata. If an image has been screenshot, re-encoded without headers, or scrubbed of metadata, no metadata signals remain to inspect. We do not claim to detect synthetic imagery through pixel-level statistical modeling.