C2PA & AI metadata — your questions, answered

Yes, it's safe: C2PACleaner reads and cleans your files entirely inside your own browser. Nothing is uploaded, stored or logged on any server — there is no server in the loop at all. Because the work is client-side, it even keeps working with your network disconnected, and your files are cleared from memory the moment you close the tab. No account, no sign-in.

It removes C2PA / Content Credentials manifests (the signed CBOR provenance blocks that AI tools embed), EXIF and GPS data, and PDF document info plus embedded XMP packets. Supported formats include JPEG, PNG and WebP images, MP4 and MOV video, and PDF documents. For writing, the Text tool strips invisible Unicode characters that can mark text as AI-generated.

It's completely free, with no account and no ads. Because everything runs locally on your own device, there's no server cost to pass on and no upload cap — the practical limit is just your device's memory. The browser tool comfortably handles everyday photos, PDFs and short videos; very large video files are better cleaned in the upcoming mobile app.

C2PA (Content Credentials) is an open provenance standard backed by Adobe, Google, OpenAI and others. It embeds a cryptographically signed record of a file's origin and edit history. AI image tools — and some cameras and editors — attach it automatically to disclose how a file was made. C2PACleaner detects that embedded manifest and removes it locally.

Platforms like Instagram (Meta) and TikTok read the C2PA / IPTC provenance metadata inside a file — along with their own signals — to add an AI label automatically. If you remove that provenance data from your own file before uploading, there's less embedded information for the automatic check to read, which can reduce an incorrect or unwanted label. It's not a guarantee: platforms may also use server-side hashes or visual classifiers that no on-device tool can affect.

Yes. Images from ChatGPT and DALL·E carry a C2PA Content Credentials manifest that marks them as AI-generated, and often EXIF fields too. Drop the image into C2PACleaner and it detects and removes that manifest and metadata on your device. It reads provenance such as the tool and timestamp — it does not expose your prompt, which generally isn't stored in the file.

Not really — and it's worth being clear about this. Removing a C2PA manifest strips one declared provenance signal, so a platform that relies on that manifest won't read it. But most AI-detection tools analyse the pixels or the text statistically and don't depend on C2PA at all, so they are unaffected. C2PACleaner is a privacy tool for controlling the metadata inside your own files — not a way to defeat detection.

Adobe Firefly and Generative Fill attach a C2PA Content Credentials manifest recording that AI was used. If you own the file and want to remove that embedded record before sharing, drop it into C2PACleaner and the manifest is stripped locally, with the pixels left untouched.

They're different layers. EXIF is unsigned camera and device data (GPS, model, timestamp) that's easy to edit. C2PA is a cryptographically signed provenance manifest about how a file was created or edited. The browser tool removes both together by default; the C2PACleaner mobile app additionally offers a location-only mode that strips just the GPS and keeps the rest of your EXIF intact.

Midjourney images generally don't embed a C2PA Content Credentials manifest today, but they can still carry basic EXIF or file metadata. C2PACleaner shows you exactly what's inside a given image before you clean it, so you can see and remove whatever provenance or EXIF data is actually present rather than guessing.

AI-written text can contain invisible Unicode characters — zero-width spaces, word joiners and bidirectional controls — plus tell-tale “smart” punctuation, which survive copy-paste and can mark writing as machine-generated. Paste your text into C2PACleaner's Text tool and it removes those hidden characters so it reads as plain text. No vendor publicly confirms a deliberate zero-width watermark; the tool simply clears whatever such characters are present.

Drop the image into C2PACleaner: before removing anything, it lists what's inside — C2PA Content Credentials, EXIF, GPS and more — so you can inspect the hidden data first. Everything is read locally in your browser, so checking an image never uploads it anywhere.

For PDFs, C2PACleaner clears the document Info fields (author, title, producer) and embedded XMP packets while keeping every page byte-for-byte identical. For MP4/MOV video, it removes container-level metadata such as creation info and any embedded provenance. As with images, the file is processed entirely on your device and never uploaded.

No. C2PACleaner is lossless — it removes only the metadata bytes and leaves the actual image data untouched, so there's no re-compression and no quality loss. The pixels you get back are identical to the original; only the hidden EXIF, GPS and C2PA data is gone.

Removing metadata from files you own is generally treated like any other edit to your own file, and it's commonly done for privacy. Stripping a C2PA manifest doesn't change who created or owns the content, and C2PACleaner is intended for protecting your own privacy — not for misrepresenting the authorship or origin of someone else's work. This is general information, not legal advice; if a specific platform's rules or a disclosure law applies to you, check those.