Skip to content
NoWatermark

Inspection tool

Remove AI Label Metadata From an Image

Some platforms read C2PA and IPTC metadata when they decide whether to label an image as AI. This finds those signals in your file, removes them, and re-scans the result so you can see what is actually gone.

Ready. Choose an image to scan.

๐Ÿ”’ Your file stays on your device.

JPG, PNG, WebP, SVG, Markdown and PDF ยท up to 25.0 MB ยท nothing is uploaded

What this changes

Some platforms read C2PA (Content Credentials) manifests and IPTC metadata, such as the DigitalSourceType field, as part of deciding whether to apply an AI label. This page finds those signals in a JPG, PNG or WebP file and removes them from your copy: the C2PA manifest, the XMP packet, IPTC records, generator tags and EXIF. By default one EXIF field, the image rotation, is kept so photos stay upright; you can turn that off.

Nothing is reported as removed because the cleaner says so. The cleaned file is scanned a second time, and a signal is marked removed only when that scan can no longer find it. The image data is copied byte for byte, so the picture itself does not change.

What no file cleaner can change

Invisible watermarks such as SynthID are shown as Unable to verify. NoWatermark cannot detect them, cannot remove them, and cannot tell you they are absent, so a clean result here is never a statement that an image will not be recognised.

Platforms also use other inputs that live outside the file: visual classifiers and their own self-disclosure rules. No file cleaner changes those. And cleaning does not affect a label that is already on a live post; it only changes the file you upload next.

Choosing what to remove

The cleanup options let you pick how much to remove. "Everything removable" drops EXIF, IPTC, XMP, Content Credentials and embedded text. "Provenance-light" drops Content Credentials, XMP and embedded text and leaves EXIF and IPTC alone. The report afterwards lists each signal as removed, kept or remaining.

Platform disclosure rules still apply to you. Removing metadata from your file does not change what a platform expects you to disclose about how an image was made, and removing provenance from content you do not own may carry legal risk in some jurisdictions.

Privacy

The scan runs entirely inside your browser. Your file is read from disk into memory, inspected there, and never uploaded โ€” you can disconnect from the internet after this page loads and everything still works.

Frequently asked questions

Will removing this metadata stop a platform from labelling my image?

It removes the metadata signals from your copy of the file, and the re-scan confirms they are gone. It cannot promise what a platform will do next, because platforms can also rely on invisible watermarks, visual classifiers and self-disclosure rules, none of which a file cleaner changes.

Does this remove SynthID?

No. SynthID is an invisible watermark in the pixels. NoWatermark cannot detect or remove it and cannot confirm it is absent, so it is always reported as Unable to verify.

My post already has an AI label. Will cleaning the file remove it?

No. Cleaning changes the file on your device, not a post that is already live. A re-uploaded cleaned file is a new upload.

Do I still have to follow the platform's AI disclosure rules?

Yes. Metadata removal does not change what a platform expects you to disclose about how an image was made. Read the rules of the platform you are posting to.

How is this different from the AI Metadata Remover?

Both use the same scan-clean-rescan engine. The AI Metadata Remover is written around prompts, model names and generator settings such as Stable Diffusion parameters. This page is written around the provenance signals platforms may read when deciding on a label.

Related tools

Related guides