Most metadata advice is written for people posting holiday photos. The stakes change when the person in the photo is a source, and the file is the only copy of something they risked something to obtain. The same tags are present; what differs is who can act on them.
This page is about process, not tools. A cleaner is one step in a chain that starts with the device that captured the file and ends with the copy a reader downloads.
Start with a threat model
Ask three questions before handling anything. Who must not learn the source's identity? What could they do with a location, a device serial or a timestamp? And who already has the file?
The answers change the plan. A photo of a public protest and a photo taken inside a private building carry the same GPS field but not the same consequence. A document whose author field names a colleague is a problem inside an organisation; the same field on a leaked internal memo is a problem for the person who leaked it.
Write the answers down. A threat model held in someone's head is forgotten at the deadline, which is when files get forwarded without inspection.
Strip before sharing
The first transmission is the disclosure. If a source sends you a photo with GPS intact over an ordinary messaging app, the coordinates have left their device and are sitting in at least two others. Cleaning the copy you publish does not recall them.
The correct order is: strip on the device that holds the original, verify the stripped copy, then send that copy. Where the source cannot do this themselves, give them a precise instruction rather than a general one — which app, which menu, what to check afterwards — and assume the instruction may be followed imperfectly.
There is a tension here worth naming. Stripping metadata also destroys provenance: the timestamps and device details that might later establish a file's authenticity. The answer is not to skip stripping. It is to keep a verified original and publish a stripped derivative, so the two purposes do not compete for the same file. The verification routine is described in verifying metadata removal.
Never rely on a platform
Some services remove metadata from images they display. Some remove only location. Some keep the original upload and serve a processed version. Some re-encode the image, which drops most tags but also changes the pixels. None of this is documented consistently, and all of it can change without notice.
Direct messages, email attachments, cloud storage links and file-transfer services are especially unreliable, because they are built to preserve the file rather than to publish it. A service that strips metadata from a public post may store the untouched original in the recipient's inbox.
The rule is simple: treat every channel as preserving everything. If the file is clean when it leaves your hands, the channel's behaviour stops mattering.
The copy you receive
A file that arrives from a source has already travelled. It passed through their phone, their messaging app, possibly a cloud backup and a laptop. Any of those steps may have preserved the original metadata, and any of them may have been observed.
So do not reason from "the file reached me" to "the file is safe to publish". Inspect what actually arrived. It may carry the source's name in an author field, their home coordinates, the serial of their phone, or the name of an internal system that generated a document.
It may also carry a trail from earlier handling: an editor's name, a resave timestamp, a thumbnail of a page that was later removed. Those are the details described in photo metadata privacy risks, applied to a file with a person attached.
| Copy | Typical risk | Handling |
|---|---|---|
| Original from the source | GPS, device serial, author name, capture time | Preserve untouched in a secure location; work from a copy. |
| Stripped working copy | Re-introduced tags from an editor's export settings | Verify the exact bytes you will publish, not the preview. |
| Screenshot | Screen content: notifications, window titles, reflections, unique objects | Treat as a photograph of a screen, not as a sanitised file. |
| Published image | Your own platform, CMS or CDN adding fields on upload | Download the live file and check it after publication. |
Keep the original separate
Store the untouched original somewhere the publishing workflow cannot reach: an encrypted volume, an offline disk, a controlled archive. Record its hash and the date it was received. That copy is for verification, legal process and editorial review, not for distribution.
Everything downstream should be derived from it deliberately. Name derivatives clearly, keep them in a different folder, and never let a "final" export overwrite the original. If a file is challenged, the preserved original lets you answer questions a stripped copy cannot.
Screenshots are not safe
Screenshots usually lose the camera's EXIF block, which makes them feel safer than they are. What they gain is everything on the screen. A notification banner, a document title, a browser tab, a filename in a window header, a reflection in a monitor — all of it is now pixels that cannot be removed by a metadata tool.
Cropping does not fix that. A crop is a new image; any identifying detail left inside the crop boundary stays. Edges are where details hide: a name badge at the frame's edge, a street sign behind a shoulder, a unique object on a desk that identifies an office.
Before publishing any screenshot, look at it as an image. Zoom in. Read every label. Then check whether the screenshot file itself carries tags added by the operating system's tool; the behaviour of common tools is covered in metadata in screenshots.
What this guide cannot do
This is general information about metadata handling, not legal advice and not security advice. It does not cover lawful process, publication risk, device compromise, network surveillance or the many operational-security questions that sit around a sensitive file. Those depend on jurisdiction, on the adversary and on facts this page cannot know.
If a source's safety depends on the outcome, involve a specialist — a lawyer, a security trainer, an editor with source-protection experience — before the file moves. Metadata hygiene is necessary and it is not sufficient. Treat it as one layer, not as a guarantee.
Within that limit, the routine is unglamorous: strip early, verify the exact file you send, distrust every platform, preserve the original, and treat screenshots as photographs. For the fields that most often survive a careless cleanup, see camera serial numbers in photo metadata.
Clean the copy first
Drop an image in. MetaPurge lists every tag, then rewrites the file without touching a pixel — in the browser, with nothing uploaded.
Open the metadata remover