Can exif.tools Metadata Prove an Image Is Authentic or Not AI-Edited?
No. exif.tools can show you what a file claims about its origin, but those claims are recorded metadata, not a verified signature. The site states this directly: origin and signing information "are recorded metadata claims, not a verified signature," and they "do not establish that the final image is authentic or free of AI edits." So metadata is evidence about the file's history, not proof of authenticity.
What exif.tools actually shows
When you load a file, the viewer reads the metadata embedded in it and displays:
- Camera & lens — camera model, lens, aperture, shutter speed, ISO, focal length
- Location — saved GPS coordinates, plotted on an OpenStreetMap view
- File details — file size, image dimensions, file type, capture time (shown as saved, without guessing a time zone)
- Origin & signing information — what the file says about its source, which may include Content Credentials-style claims
All of this is read locally in your browser. The site notes your files stay on your device, and links you paste load directly with a proxy fallback if needed.
Why metadata can't prove authenticity
Metadata fields are just text values written into the file. Anyone can edit them with the right tools, and editing them doesn't change the pixels. A camera tag naming a specific phone model, for example, is a claim the file makes — it isn't a cryptographic guarantee that the sensor data came from that device.
The site is explicit that these origin records "can include source ingredients and do not establish that the final image is authentic or free of AI edits." It also points you to verify Content Credentials externally rather than treating the in-file claims as final.
The SynthID problem: watermarks live in pixels, not metadata
Google's SynthID watermark is embedded in the image pixels themselves. exif.tools notes that ExifTool-style metadata inspection cannot detect it — the watermark isn't a metadata field, so a metadata viewer will never see it.
The site's suggested path: open Gemini, attach the image yourself, and ask it to check for a SynthID watermark. There's a direct link to Google's SynthID check in Gemini.
One critical caveat the site states plainly: a missing watermark does not prove an image is real. Many real photos were never watermarked, and watermarks can be removed or degraded by editing, resizing, or re-encoding.
What you can and can't conclude
| What you find | What it means |
|---|---|
| Camera, lens, ISO, aperture present | The file contains camera-style metadata — consistent with a camera capture, but editable |
| GPS coordinates present | A location was saved in the file; useful for context, not proof of where the photo was taken |
| Origin/signing claims present | The file claims a source; not a verified signature |
| No SynthID watermark detected | Inconclusive — absence of a watermark proves nothing about authenticity |
| Metadata stripped or absent | Common after social media uploads and messaging apps; tells you little on its own |
A practical way to use this
Treat exif.tools as a first-pass inspection, not a verdict:
- Load the file and check whether camera settings and capture time are present and internally consistent.
- Look at origin and signing claims, but read them as claims — then verify Content Credentials through the external route the site links to.
- If AI-origin detection matters, run the SynthID check in Gemini separately, since metadata inspection can't see it.
- Combine what you find with other evidence — reverse image search, source context, pixel-level forensics — before drawing conclusions.
The honest summary: metadata tells you what a file says about itself. Proving an image is authentic, or proving it wasn't AI-edited, takes more than that.