Why platforms add an AI label
Image files can contain more than the picture you see. Content Credentials (C2PA) can record how an image was created or edited. The IPTC digital source type can declare AI use. This tool reads those records and shows you the metadata present in your file.
Removing these records produces a copy without that metadata. It does not remove invisible watermarks or control a platform’s own detection. Explore the platform guides for more context, or use the AI label checker to inspect an image first.
How it works
Check
Drop an image. It is parsed on your device: EXIF, XMP, IPTC, PNG text chunks and any C2PA manifest. You get a verdict and the exact evidence behind it.
Remove
Strip the metadata losslessly, so pixels stay byte-identical, or re-encode into a fresh JPEG, WebP or PNG with an optional gentle resize.
Verify and download
The output is parsed again before a download button appears. If anything removable were left behind, you would see it.
Two ways to remove it
Strip metadata (lossless)
- Removes Content Credentials, XMP, IPTC, EXIF, PNG text and comments.
- Keeps the encoded pixels byte for byte. No quality loss, ever.
- Keeps the ICC color profile and the rotation flag so the image looks identical.
- Best for photographs, print work and anything where quality matters.
- The file is parsed again after cleaning, and the download only appears if the re-check passes.
Re-encode and resize
- Decodes the image and writes a brand-new JPEG, WebP or PNG with zero metadata.
- Optional gentle shrink (3%, 10% or a fixed width) changes every pixel slightly.
- Breaks the hash that binds a C2PA manifest to the pixels and defeats exact pixel matching.
- Quality 90 to 95 is visually lossless for social feeds. Transparency is kept for WebP and PNG.
- The right choice when a file cannot be cleaned losslessly, or when you want a fresh file anyway.
Which platforms read this metadata
Scroll sideways to see all columns →
| Platform | What it reads | What appears | Guide |
|---|---|---|---|
| Instagram, Facebook, Threads | C2PA Content Credentials and IPTC digital source type, plus invisible watermarks from partner generators | “AI info” on the post, or tucked into the post menu when only edits were AI | Instagram guide |
| TikTok | C2PA Content Credentials, plus its own detection and creator disclosure | “AI-generated” label; realistic AI content without a label can be removed | TikTok guide |
| C2PA Content Credentials | A “CR” icon on the image that opens the recorded history, including AI use | LinkedIn guide | |
| IPTC and C2PA metadata, plus classifiers | “AI modified” label; people can choose to see fewer AI Pins | Pinterest guide | |
| Google Search and Photos | IPTC and C2PA metadata, plus the SynthID watermark in Google-made images | “AI info” in Google Photos and “About this image” in Search | Metadata explained |
| X | Little to none; metadata is discarded on upload | No metadata-based label |
What removing metadata cannot do
- Invisible watermarks. Google’s SynthID, Meta’s own watermark and similar systems live in the pixels, not the metadata. Stripping metadata does not touch them, and a gentle resize does not reliably remove them either.
- Platform detection. Meta, TikTok and Pinterest also run their own classifiers and honor self-disclosure. A clean file removes the most common trigger; it does not guarantee that no label will appear.
- Platform rules. Most platforms require you to disclose realistic AI-generated content. Use this to stop false labels on real photos, to remove private data, and to clean images you are allowed to publish. Do not use it to pass off synthetic media as real.
Your images are processed inside your browser, with no upload endpoint and no account. Checking, cleaning and converting all happen on your device. Read how your privacy is protected.
Questions people ask
Does this remove the “AI info” label on Instagram?
It removes the metadata Instagram reads. When the label was triggered by C2PA or IPTC metadata, which is the usual case for images from Photoshop, Firefly, DALL-E or Midjourney, a cleaned upload normally does not get the automatic label. If the label comes from an invisible watermark or from Meta’s own detection, it can still appear. See the Instagram guide for details.
Does stripping metadata lower image quality?
No. The lossless path copies the encoded image data byte for byte and only drops the metadata segments. Color profile and rotation are kept. Re-encoding is a separate option that does decode and re-compress the image.
Why would I re-encode or resize instead?
Content Credentials bind a manifest to a hash of the pixels, and some systems match pixel fingerprints rather than metadata. Re-encoding at quality 92 and shrinking by 3% produces a new file with no metadata and slightly different pixels, with no visible loss. It is also the fallback when a file cannot be cleaned losslessly.
Which formats are supported?
JPEG, PNG and WebP can be inspected and stripped losslessly. Re-encoding outputs JPEG, WebP or PNG. HEIC and AVIF are not supported yet; export them as JPEG first.
Is it allowed to remove this metadata?
Removing metadata from files you own or have the rights to is a normal, legal operation, and platforms strip most of it themselves on upload. Platform rules still require disclosure of realistic AI-generated content, so use the tool to correct false labels, protect privacy and clean files you are entitled to publish.
Are my photos uploaded anywhere?
No. The site has no server-side processing. Files are read by a Web Worker in your browser, the cleaned copy is created in memory, and closing the tab discards everything.