Does EXIF data prove that a photo is real?
No. EXIF is plain, unsigned text that any free tool can create, edit or copy, so its presence does not prove a camera took the picture and its absence does not prove an image is AI. It is a useful clue when its fields agree with each other. A signed record such as C2PA is stronger, and the original file is stronger still.
Last updated: 2026-10-08
What EXIF is
EXIF is a block of tagged values stored in a JPEG or a TIFF-based file. It records what the device chose to write at the moment it saved the file: the make and model, the date and time, the exposure settings, the lens, the orientation and, when enabled, a GPS position. Editing programs read it and write it back, and many add their own name to it.
The format has no signature. Nothing binds the values to the picture. A program that opens the file can change any field and save it, and the next reader has no way to know.
What it can suggest
A file with a coherent set of EXIF fields is more likely to have come from a camera than a file with none, and the difference is real. The fields interact in ways that are hard to fake without care.
- The lens and the focal length should agree with the device.
- The exposure triple should suit the scene and the time of day.
- The date should be plausible in relation to the file's other timestamps.
- A thumbnail should show the same picture as the main image.
- A maker note, a binary block of model-specific data, should be present and well-formed on cameras that write one.
When these line up, the metadata is a good clue. When they clash, such as an aperture the lens cannot reach, it is a flag worth following.
What it cannot prove
Presence proves nothing, because the fields can be written in by hand. Absence proves nothing as well. Platforms remove EXIF when you upload for privacy reasons, messaging apps do the same, and a screenshot never had any. A photograph that has passed through two or three apps will usually have none.
So the clean statement is this: EXIF is evidence about a file's history only when you can trust the chain it came through, and you almost never can.
Stronger signals
Several kinds of evidence carry more weight than EXIF.
1. The original. A file straight from the device, with a modification date that matches the capture, is hard to dismiss. 2. A RAW file, or neighbouring frames from the same shoot, which show the scene before and after and which a generator does not produce. 3. A signed provenance record. C2PA binds a signed claim to a hash of the content, so editing the pixels breaks the match. The guide on what C2PA is explains the design and its limits: it tells you who signed a claim, and it still cannot tell you the claim is true. 4. A pixel watermark such as SynthID, for the tools that write one. That is a mark in the pixels, described in how SynthID works.
None of these is universal. Most photographs in circulation have none of the last two.
Practical advice
If you are reviewing an image, read the EXIF as a prompt for questions and not as an answer. Check that its fields agree. Ask for the original. Be careful with a detector score, because those are estimates, as the page on why detectors flag real photos explains.
If you are publishing an image, expect readers to treat metadata the same way. It is fine to keep your camera EXIF on a real photograph, and it is useful to keep the original. If you add camera fields to a generated file, you are writing a claim, so do it consistently, and be ready to say how the image was made when someone asks. PassReal writes a consistent device block as part of its encoding, and it is a convenience for the file, not a certificate of origin.
Questions
Can EXIF be faked?
Yes, easily. ExifTool and many editors write any value into any field, and copying a whole block from one file to another takes one command.
Does missing EXIF mean the image is AI?
No. Social networks, messaging apps and screenshots all remove EXIF from genuine photographs, so a missing block is common in real images.
What would be stronger evidence than EXIF?
The original file from the device, a RAW file, other frames from the same shoot, and a signed Content Credentials record where the camera produces one.
Can an AI image have real-looking EXIF?
Yes. A file can carry a camera make, model, date and exposure that someone wrote in, which is why a reviewer should check that the fields agree with each other and with the picture.
Do platforms treat EXIF as proof?
Platforms rely more on standardized provenance signals such as C2PA and the IPTC digital source type, and on their own analysis, than on camera fields.
Limits
Naturalization operates on the pixels and on the file, not on the content of the image. A subject that is implausible - six fingers, inconsistent reflections, text that does not read - stays implausible after processing, and a human reviewer will notice it. PassReal changes what a statistical classifier measures, not what a person sees.
Results vary by generator, by subject and by detector, and detectors are retrained. No pass rate is published on this page because none has been measured in a way that would still hold next month.
Three images, no card, no expiry.