AI Image Detector guide
Inspect an image for embedded AI-generator markers, camera metadata, and provenance clues without uploading the file.
What this tool does
The AI Image Detector examines file-level evidence in JPG, PNG, and WebP images. It searches supported metadata and embedded text for recognizable generator or workflow names such as Stable Diffusion, Midjourney, DALL-E, Adobe Firefly, ComfyUI, and similar systems. It also reports camera-related EXIF fields, editing-software labels, and possible Content Credentials markers. A short SHA-256 prefix helps distinguish the exact file that was inspected.
Results use four careful outcomes: AI-generation signals found, camera workflow clues found, conflicting origin signals, or no decisive origin evidence. The checker deliberately avoids a fabricated probability score. Current research and real-world testing do not support reliable universal detection from image pixels alone, especially after screenshots, resizing, recompression, editing, or metadata removal.
An explicit generator label is useful evidence, but metadata can be removed, edited, or copied into another file. Camera tags also are not proof that every visible pixel came directly from a camera. An image can combine photography, conventional editing, generative fill, and fully generated elements in one composition.
How to use it
- Select a JPG, PNG, or WebP file up to 25 MB for metadata inspection.
- Select Check origin clues.
- Read the evidence result and its explanation.
- Review each generator, camera, editing, provenance, or limitation signal.
- Keep the exact file available if the result will be reviewed by another person.
For important verification, combine this result with the original source, publication history, reverse-image search, signed Content Credentials verification, and direct confirmation from the creator or publisher. Do not accuse a person of deception based only on this tool.
Use the Image Metadata Viewer for a broader list of supported EXIF and container details. If disclosure reduction is the goal, create a separate copy with the Image Metadata Remover and verify it independently.
Benefits
- Identifies recognizable generator and workflow metadata
- Reports supported camera and editing-software fields
- Flags possible provenance containers without claiming validation
- Uses clear evidence categories instead of misleading percentages
- Fingerprints the inspected file with a SHA-256 prefix
- Keeps image reading and analysis on the current device
Understanding limitations
Most social platforms remove metadata, and screenshots normally contain metadata from the screenshot device rather than the original image. A fully AI-generated picture can therefore produce “no decisive origin evidence.” A photographed screen displaying generated artwork can contain genuine camera metadata. Conversely, a human-made image may pass through software whose name resembles an AI workflow.
This checker does not analyze invisible watermarks, verify C2PA signatures, search the web, identify a creator, determine copyright ownership, or provide forensic, legal, academic, or journalistic proof. Its role is to surface inspectable clues and explain their limits.