EXIF, C2PA Content Credentials and reverse image search are frequently grouped as “image verification,” but they observe different layers. EXIF describes properties stored in one file. C2PA can carry signed assertions about provenance and edits. Reverse search compares visual appearance with indexed web images. A strong investigation uses all available layers without asking any one of them to prove more than it can.
A realistic file journey
Consider one camera photo that is exported from an editor, posted to a social network, downloaded, placed in a message and finally captured as a screenshot. The visual scene may remain recognisable while evidence changes at every step. The original may contain GPS and camera time. The edited export may retain some EXIF but change dimensions and software fields. The platform copy may remove EXIF and provenance. The screenshot creates an entirely new image whose metadata describes the screenshot device, not the original camera.
| Layer | Useful question | Unsafe conclusion |
|---|---|---|
| EXIF | What does this file claim about capture and device? | GPS exists, therefore location is verified |
| C2PA | Is a signed manifest intact and what actions are recorded? | No credentials, therefore fake or human-made |
| Reverse search | Where have visually similar versions appeared? | Oldest indexed page is the original source |
| Visual geography | Does the scene agree with the claimed place? | Similar architecture proves an exact street |
Test EXIF as a claim, not an oracle
Preserve the file and calculate a hash before editing. Extract GPS, DateTimeOriginal, timezone offset when present, camera model, software, orientation and dimensions. Record missing fields explicitly. Then ask whether the fields are internally consistent: does the timezone fit the coordinate; does editing software explain a later modification time; do orientation and dimensions match the displayed image?
Cross-check the scene. A GPS point in London is suspicious if the image contains unambiguous tropical geography; a one-hour time discrepancy may simply reflect daylight saving. Metadata inconsistency lowers confidence but does not automatically prove manipulation.
Verify Content Credentials precisely
C2PA manifests can contain signed claims about creation, editing and ingredients. Verification establishes whether the manifest and signatures validate under the standard. It does not guarantee the camera was physically at a claimed location unless that assertion exists, is trusted and agrees with external evidence.
Absence is weak evidence. Credentials can be lost through screenshots, unsupported editors, exports and platforms. Record “no manifest found in this file,” not “this image has no provenance.” The distinction matters because you are examining one derivative.
Use reverse search to reconstruct publication history
- Search the full image.
- Search distinctive crops that survive edits.
- Compare dimensions and crop boundaries.
- Open the actual pages and record publication dates, not search snippets alone.
- Look for a higher-resolution version and photographer credit.
- Distinguish first indexed, earliest observed and original creator.
A matching news page proves that the page used the image. It does not prove the caption, date or capture location. Several websites may repeat the same false description from a syndication feed.
Build an evidence matrix
Place each claim—creator, capture time, location, editing history, first publication—in a row. Put EXIF, credentials, web history and visual verification in columns. Mark support, contradiction or no evidence. A conclusion becomes strong when independent columns agree; repetition within one column is not independence.
Minimum defensible language
- “The file contains GPS coordinates consistent with the visible scene.”
- “The C2PA manifest validates and records these edits.”
- “No manifest was found in the examined derivative.”
- “This is the earliest version located in the searches performed.”
- “The publication location is not the capture location.”
Official references
- C2PA Technical Specification
- OpenAI: C2PA and why metadata may disappear
- Google Photos: how a photo was made
Put the method into practice
When you have a photo to investigate, start with our AI photo location finder and treat its result as a lead to verify, not automatic proof.