The pile inside a photo
A JPEG, PNG or WebP can hold more than pixels. The extra records are written by cameras, phones, editors, and sometimes generative tools. They are not a diary of everything that ever happened to the picture.
The useful split is: EXIF for camera settings and often time; GPS tags when location was stored; XMP for editor and rights fields; IPTC for captions and contacts; C2PA for signed provenance claims.
What is usually not in “the metadata”
A face in the frame, a street sign, or a Discord username in a screenshot is not EXIF. Clearing headers does not blur those pixels.
Invisible pixel watermarks such as SynthID are also not metadata. A header strip does not remove them. Absence of C2PA is not proof the image is a camera original.
Why the same photo looks different after an app
Instagram, Discord, WhatsApp and iMessage often recompress. That can drop GPS, keep a timestamp, or replace the file entirely. The original on disk and the downloaded post are different objects.
If you need a fact, inspect the file you still control — the export, the camera roll original, or the attachment — not a screenshot of the feed.
A local check beats guessing
You can read supported headers in the browser with Metadata Cleaner. Nothing about that step requires uploading the media. If the question is Content Credentials or generator signals, use AI Image Origin Detector for local inspection first.