C2PA Content Credentials
Provenance manifests can describe how content was created or edited. They are not inherently proof of AI generation.
Free Instagram AI Info rescue
Instagram added “AI Info” to your photo? Clean the file signals before you post again.
Scan supported C2PA, IPTC, XMP, EXIF, GPS and editor traces in your browser, remove the fields that can be cleaned and verify the new copy before download. Then let AI prepare a reviewed context copy, alt text and post ideas from a reduced metadata-free preview.
Free · no account · original stays on your device · reviewed July 19, 2026
What Meta says
AI Info is broader than one EXIF field.
Meta says it can label content when it detects industry-shared AI signals or when a person self-discloses. Its published approach also discusses invisible watermarking and classifiers. In a later update, Meta acknowledged that the old “Made with AI” wording did not always match people's expectations for images that used only minor AI edits, and moved to “AI Info” language.
Primary sources: Meta's labeling announcement and Meta's policy update. Platform behavior can change; these links are the authority, not this tool.
Signals inside the file
Four structures worth inspecting separately.
XMP and EXIF software
Editors and generators can leave software, workflow, creator and rights information in these containers.
PNG text chunks
Some generation workflows store prompts, seeds, samplers, models or workflow notes as ordinary PNG text.
Outside ordinary metadata
A cleaner cannot see or remove every platform signal.
- Invisible watermarks Signals encoded into pixels or recovered through durable-credential systems can survive ordinary metadata removal.
- Visual classifiers A platform can analyze the image itself; byte-level metadata cleaning does not change what the picture depicts.
- Post-level records A label already attached to a post, account history or a creator's disclosure lives on the platform, not in your local file.
Responsible workflow
Inspect → choose → clean → verify.
- Select a file you own or are authorized to process; it stays in browser memory.
- Review privacy fields and provenance fields as different kinds of information.
- Use Standard for a quality-guarded copy, or Inspector & batch when you need exact coverage.
- Download the cleaned copy and review the before/after report before publishing.
- Disclose synthetic or materially altered media wherever platform rules or applicable law require it.
No “undetectable” promise.
This is a privacy and provenance-control tool, not an AI-detection bypass guarantee. A metadata cleaner cannot prove an image's origin, raise its reach, remove a label already applied, or erase watermark and classifier signals outside supported file structures.
AI Info FAQ
Three boundaries to remember.
Can metadata cleaning remove Instagram's AI Info label?
It can remove supported embedded signals that may be one input before upload. It cannot undo a label already applied or remove signals outside ordinary metadata, so no outcome is guaranteed.
Does C2PA or XMP prove that an image was generated by AI?
No. C2PA is a general provenance standard and XMP is a general metadata container. A reliable report distinguishes the container from an explicit AI source or workflow assertion.
Does a clean metadata report prove a photo is not AI-generated?
No. It only reports that this inspector did not find supported fields. Files routinely lose metadata through normal export, messaging and editing workflows.