Generator · pre-upload
Cleaning C2PA from Meta AI images
Images from Meta AI carry C2PA, and posting them to Meta's own platforms means that C2PA is read first. The combination makes a label close to inevitable.
On-device/No upload/Quality unchanged/No sign-up
Drop images here, or click to choose
JPG, PNG, WebP, AVIF, HEIC · 15 MB each · up to 30 per batch
You can also paste a screenshot with ⌘V / Ctrl+V
Cleaning options
Want to look first? Themetadata inspectorlists every field in the file, so you can see exactly what is there before deciding what to strip.
Mostly C2PA, and Meta platforms read it first
Different export paths can write different things. The reliable approach is to inspect one sample before batch processing.
Where it writes
| Location | Notes |
|---|---|
| C2PA | Content credentials written at generation |
| Platform reading | Instagram / Facebook check C2PA first |
| XMP | May name the model |
Pre-upload checklist
- 11. Clean C2PA — this is the key one
- 22. Clear XMP as well
- 33. Verify again before posting
- 44. Upload that copy, with no editor or forwarding in between
Common traps
- Editing again inside a Meta app, which may write it back
- Clearing XMP while leaving C2PA in place
If it is still flagged after cleaning
That means the trigger was never in the file metadata. It could be an invisible pixel-level watermark, the platform's own visual classifier, or history held on the platform side. None of those live in the metadata layer, and cleaning again will not change them.
The full boundaries are in thedisclaimer; if you just want to confirm the file really is clean, use thewatermark check.
Common questions
Why is Meta AI output flagged so readily?
Because the generator and the judge are the same company, and C2PA sits at the top of its reading order.
Does clearing C2PA actually help?
For metadata-driven decisions, yes.
Does it affect quality?
No.