Skip to content
Holiday saleSave up to $81 on an annual package, from $3.75/mo.Ends in--d --h --m --sGet the deal

Remove C2PA Content Credentials From Images and Video

C2PA Content Credentials are the signed record inside a file that tells Instagram, TikTok, YouTube and LinkedIn an image or video was made or edited with AI. StripShot removes that record at the binary level, in your browser, with no re-encoding. This page explains where the credential lives, who reads it, what removing it changes, and what it cannot change.

Videos save best in Safari or Chrome.

0/2 images today

What a Content Credential is

C2PA is the open specification behind "Content Credentials". A credential is a manifest: a signed set of statements about the file (which app made it, whether generative AI was involved, what edits were applied) plus a cryptographic hash of the content so later tampering can be detected. The C2PA 2.1 specification calls that hash a hard binding.

Apps that write credentials today include Adobe Photoshop, Lightroom and Firefly, OpenAI's image tools, Microsoft Designer, Google's Pixel camera and several AI video generators. The IPTC DigitalSourceType field often travels with it: trainedAlgorithmicMedia means "created using generative AI" and compositeWithTrainedAlgorithmicMedia means "edited using generative AI". Platforms read both.

Where the credential is stored

File typeContainer (per the C2PA spec)What StripShot removes
JPEGAPP11 marker segment (JUMBF box)The APP11 segment(s), plus XMP and IPTC provenance fields
PNGcaBX chunkThe caBX chunk, plus tEXt/iTXt chunks with AI parameters
MP4 / MOVuuid box before the first mdatThe uuid box, the XMP block and software atoms (©too, ©swr)
WebP, HEICMetadata chunks / boxesEXIF and XMP metadata

Because the credential sits in its own container next to the image data, it can be removed without decoding the pixels. That is why StripShot's output is bit-identical in the picture itself; tools that "clean" by re-saving through a canvas recompress the image and lose quality.

Which platforms read it, in their own words

Instagram, Facebook, Threads

Meta applies its "AI info" label when it detects "industry standard AI image indicators", naming C2PA and IPTC (Meta, Feb 2024), or when the poster discloses. The label was renamed from "Made with AI" on July 1, 2024, and AI-edited content gets it in the post menu only (Meta, Apr 2024, updated Oct 2025).

TikTok

Reads Content Credentials on uploaded images and video since May 2024 and auto-labels them "AI-generated". Also uses "our own detection models" and, since Nov 2025, is testing invisible watermarks on content made with its own AI tools. TikTok, May 2024, Nov 2025.

YouTube

Requires creators to disclose realistic altered or synthetic content, and since May 2026 may add the label automatically from its own AI tools, from C2PA metadata, or from content "our internal systems detect is AI generated". Penalties for not disclosing include removal and suspension from the Partner Program. YouTube Help, May 2026.

LinkedIn

Shows a C2PA icon on image and video content that carries a signed credential; tapping it reveals the app, the issuer and any AI-use statement. No credential, no icon. LinkedIn Help.

Pinterest

Its "AI modified" label (Apr 2025) checks metadata against the IPTC standard (Pinterest help) and also uses "classifiers that automatically detect Gen AI content, even if the content doesn't have obvious markers" (Pinterest, Apr 2025). Pinterest does not name C2PA.

Not published

Whether Instagram strips or keeps metadata on upload, whether Meta runs a pixel detector, and X's current label rules: none of these have a public statement from the platform. Treat anyone who states them as fact with caution. Full rundown in our platform label tracker.

How to remove C2PA Content Credentials

  1. 1Check what the file contains. Drop the image or video into StripShot. The scanner lists any C2PA manifest, XMP, IPTC and EXIF it finds, read-only, before anything changes.
  2. 2Strip the Content Credentials. Click strip. The C2PA container (APP11 in JPEG, caBX in PNG, the uuid box in MP4/MOV) is removed along with XMP and IPTC provenance fields. Pixels and video frames are untouched.
  3. 3Verify. StripShot shows a SHA-256 check of the image data before and after. You can also drop the clean file into the free checker, or into the Content Authenticity Initiative's Verify tool, to confirm no credential remains.
  4. 4Upload the clean copy. Post the cleaned file. A label already applied to a published post cannot be removed afterwards; delete and re-post the clean file.

Want to see it before you trust it? Drop any file into the free AI metadata checker, which lists the credential without changing anything, and verify the cleaned copy at the Content Authenticity Initiative's Verify tool: it should report no Content Credentials.

What removal changes, and what it does not

It changes: the file carries no C2PA manifest, no IPTC digital-source field and no XMP tool signature. A platform whose label logic reads those signals, which is what Meta and LinkedIn describe, has nothing left to read. The Content Authenticity Initiative itself notes that "metadata of any kind can be removed deliberately or accidentally" and that a screenshot "effectively remove[s] secure metadata" (CAI, Apr 2024). Binary removal does the same thing without the quality loss of a screenshot.

It does not change: anything in the pixels. Google's SynthID is "embedded in the pixels of an image", designed for cases where "metadata can be manually removed or even lost when files are edited" (Google DeepMind, 2023), and now covers images, audio, text and video from Gemini, Imagen, Veo and Lyria (SynthID overview). Platform classifiers that look at the frames themselves (TikTok, Pinterest and YouTube all say they run them) are also unaffected. And C2PA's soft binding lets a watermark or fingerprint re-link a file to a copy of its manifest if one was published to a provenance store; that is rare today, but it exists in the spec. No metadata tool, StripShot included, can promise a label will never appear.

Who adds credentials without telling you

The most common surprise is a real photograph labeled as AI. Adobe Photoshop and Lightroom attach Content Credentials when you use features such as Generative Fill or Remove, and the credential records that generative AI touched the file. Meta's own update of Sep 12, 2024 moved the label for AI-edited content into the post menu precisely because of these cases. If you want a real photo to stay unlabeled, strip the credential before posting, or avoid generative features in the edit.

The rules around it

There is no general law against editing metadata in a file you own. Three specifics matter. Adobe's Generative AI User Guidelines (May 2026) state "You must not remove, alter, or disable any Content Credentials" on Adobe generative output. The EU AI Act, Article 50, in force from August 2, 2026, requires AI-generated or manipulated content to be "marked in a machine-readable format and detectable" and requires deepfakes to be disclosed. California's AI Transparency Act (SB 942, amended by AB 853) also took effect August 2, 2026 for covered providers. If you publish AI-generated content to an audience in those places, disclosure is on you regardless of what the file carries. Our disclaimer has the plain version.

Last verified against the sources linked above on October 30, 2026. Platform rules change; the tracker is updated monthly.

Pricing

Clean 2 images free. Then go unlimited.

No card needed to start. Upgrade anytime.

Holiday Sale, through November 30: a first year of Pro for $45, then $84/yr, or Full Access for $99, then $180/yr. See the Holiday Sale

Free
$0
2 images/day free
  • 2 images/day free
  • 1 video every 30 days
  • AI detection (70+ signatures)
  • 100% private, no uploads
Clean your image free
POPULAR
Pro
$9/mo
Unlimited photos
  • Unlimited image cleanings
  • 10 videos/day included
  • Batch processing
  • Smart Strip & hex viewer
  • Edit & write EXIF tags
Start Pro for $1$1 for 14 days, then $9/mo. Cancel anytime.Or get annual for $3.75/mo, billed yearly. Ends Nov 30.
Full Access
$19/mo
Unlimited everything
  • Unlimited image cleanings
  • Unlimited video cleanings
  • Strip Sora, Runway, Kling
  • MP4 + MOV support
  • New generators added regularly
Start Full Access for $1$1 for 14 days, then $19/mo. Cancel anytime.Or get annual for $8.25/mo, billed yearly. Ends Nov 30.
Cancel anytime  ·  No commitment  ·  Files never leave your device

Last updated October 30, 2026. Platform rules change; see the platform label tracker for what each platform currently reads.

Frequently asked questions

What are C2PA Content Credentials?

C2PA is an open standard from the Coalition for Content Provenance and Authenticity (Adobe, Microsoft, Google, OpenAI, Intel and others). A Content Credential is a signed record stored inside the file that says which app created or edited it and whether generative AI was used. Platforms read it to decide whether to show an AI label.

Where is C2PA data stored in a file?

In a JPEG it is an APP11 marker segment. In a PNG it is a caBX chunk. In MP4 and MOV video it is a uuid box placed before the first mdat box. StripShot finds and removes exactly those containers; the image or video data is not touched.

Does removing C2PA Content Credentials stop the Instagram AI label?

Meta says its label is applied from industry-standard signals in the file (it names C2PA and IPTC) or from the creator's own disclosure. With those signals removed, a photo carries nothing for Meta's systems to read. Meta has not published whether it also runs a detector on the pixels, so no tool can promise a label will never appear.

Does StripShot remove C2PA from video?

Yes. MP4 and MOV from Sora, Runway, Veo, Pika and others are cleaned at the container level: the C2PA uuid box, the XMP block and the software atoms are removed without re-encoding the video, so there is no quality loss.

Can a platform still tell an image was AI-made after the credentials are gone?

Sometimes. Google's SynthID is written into the pixels of images and video from Gemini, Imagen and Veo, and Google says it survives metadata removal. Pinterest and TikTok say they also run their own detection models. C2PA also allows a soft binding (a watermark or fingerprint) that can re-link a file to a copy of its manifest if one was stored elsewhere. StripShot removes the metadata; it does not change pixels.

Does removing Content Credentials reduce image quality?

No. StripShot removes the metadata containers at the binary level and never decodes or re-encodes the pixels. After every strip it shows a SHA-256 check so you can confirm the picture itself is unchanged.

Is it legal to remove C2PA Content Credentials?

Editing metadata in a file you own is normal; there is no general law against it. Two things to know: Adobe's generative AI user guidelines say you must not remove or alter Content Credentials on Adobe generative output, and from August 2, 2026 the EU AI Act (Article 50) requires AI-generated content to be marked and detectable, with similar rules in California's AI Transparency Act. How you publish is your responsibility; this page is not legal advice.

Is my image uploaded anywhere?

No. Scanning and removal run inside your browser. The file never leaves your device.

Related

AI label remover

Remove the "AI info" label from photos.

Remove AI metadata from video

MP4 and MOV from Sora, Runway, Veo, Pika.

Remove C2PA from a video

Content Credentials in MP4 and MOV files.

Check a file for C2PA

Read-only: see what a platform will see.

Glossary: C2PA

Manifest, hard binding, soft binding.