How to check C2PA Content Credentials in an image
A Content Credentials verify tool shows the signed manifest: issuer, actions such as created or edited, and the tools named on those actions. It does not show social download counts. Width and height can sit in an EXIF assertion, or you read them from the file. A missing manifest is inconclusive.
How to check C2PA Content Credentials in an image?
A Content Credentials verify tool shows the signed manifest: issuer, actions such as created or edited, and the tools named on those actions. It does not show social download counts. Width and height can sit in an EXIF assertion, or you read them from the file.
C2PA: Coalition for Content Provenance and Authenticity; C2PA specification 2.4; Instagram Help: label AI content before sharing; YouTube Help: disclosing use of GenAI content
Last updated
Keep the original file. For every check, record the detector or product name, its version or the date you checked, the supported scope that product currently states, the result text exactly as shown, and any transformations already applied to the file (crop, resize, screenshot, re-encode, or another export).
C2PA is an open provenance standard used to attach signed claims about how media was created or changed.
A file can be inspected for a C2PA manifest, signer information, assertions, and whether the credential validates.
No credential does not mean “not AI.” Metadata can be absent, stripped, unsupported, or replaced by another provenance system.
What C2PA actually is
C2PA stands for the Coalition for Content Provenance and Authenticity. The C2PA 2.2 explainer describes signed provenance claims bound to content, plus a trust model for who signed them. The visible picture is still the picture; the credential is a structured record associated with it.
That record can describe the tool that created or exported a file, editing actions, ingredients used in a composition, timestamps, and other assertions. The exact fields depend on the software and workflow. C2PA is not limited to AI. A camera can use provenance credentials, a newsroom can use them, and an AI image generator can use them.
Keep four questions apart:
- Presence. A container or manifest is there, and fields can be read.
- Signature and content binding. The credential verifies against the bytes it claims to cover.
- Trust. The verifier’s policy accepts the signer or issuer.
- Truth. The depicted events or asserted claims are actually true.
Readable fields are not cryptographic validation. A verified credential is still not factual truth about what the picture shows. Absence of a credential is inconclusive: the file may never have had one, it may have been stripped, or a different provenance system may be in use.
What a Content Credentials verify tool shows
When you use a Content Credentials verify tool on a file that still has a manifest, the record can include the issuer, editing actions, and the software or tools named on those actions. The C2PA 2.4 specification defines actions and optional assertions such as EXIF. Ingredients appear when the manifest lists them. Social-media download counts, likes, and share totals are not fields in that manifest. Image width and height can appear inside an EXIF assertion if the signer stored one, and you can also read the pixel size by opening the file. File size is the size of the bytes you have, not a download statistic. A missing manifest is inconclusive: the file may never have had one, or a later save may have dropped it.
On this site, drop the original image in the C2PA viewer or the Content Credentials checker. Choose image. Read the signer and the actions. Optional Verify with OpenAI is a separate click and sends the file. The definition of the note stays on what Content Credentials are.
How to check a file
- Keep the original file. Do not begin by taking a screenshot or re-saving it through another app.
- Name the verifier. Write the product, version or check date, and the supported scope it currently states.
- Copy the result verbatim. Save the exact on-screen wording, not a paraphrase, and list transformations already applied to the file.
- Read the claims, not just the badge. “Credential present” is less useful than who signed it and what the history says.
- Check for other provenance signals. Pixel-domain watermarks such as SynthID are a separate mechanism from C2PA metadata.
- Interpret a negative result narrowly. “No supported signal detected” is not the same conclusion as “this image was definitely not AI-generated.”
C2PA and SynthID are different layers
| Signal | Where it lives | What it can help establish | Main limitation |
|---|---|---|---|
| C2PA / Content Credentials | Structured signed provenance associated with the media file | Origin and history claims from an identified signer | Credentials can be absent or lost during some workflows |
| SynthID | Imperceptible watermark embedded in generated content | Whether supported content carries a watermark associated with a participating generator | Detection is system-specific and not universal across every AI model |
| Ordinary EXIF/XMP | File metadata | Camera, software, timestamps, GPS, authoring fields | Usually not cryptographically trustworthy by itself |
| Visual classifier | Model inference over pixels | Probability-like evidence of synthetic visual patterns | Can produce false positives and false negatives |
OpenAI and Google expose separate verification products
OpenAI’s current verification flow checks supported files for provenance signals associated with OpenAI tools, including trusted C2PA manifests and supported SynthID watermarks. A positive result can tell you that the file contains an OpenAI-associated provenance signal. It does not tell you that the image is factually correct, unedited, or owned by a particular person.
CreatorPrivacyKit’s AI Image Origin Detector inspects supported container structures locally after Choose image. That local pass does not upload the file. Optional Verify with OpenAI is an explicit external check: it sends the selected image to OpenAI. Download JSON report saves the local report. CreatorPrivacyKit does not run Gemini’s SynthID check, Gemini’s Content Credentials facility, or Google DeepMind’s independent SynthID Detector.
Google’s products are not one detector. Gemini’s in-product SynthID check is separate from Gemini’s Content Credentials facility, and both are separate from the independent SynthID Detector. Gemini help accessed 2026-10-09 says the Gemini SynthID check recognizes Google AI content. That Gemini-help scope is not a blanket limit for every SynthID tool.
The same Gemini help uses negative phrasing in its verification-results language: a missing watermark can indicate the content is not from Google AI tools that embed SynthID. It also states the alteration limits that keep that reading bounded: watermarks can be missed after many transformations, and results can be unclear for simple or abstract content or for tiny edits. Those conditions stay attached to the negative phrasing. A Gemini “not found” result is not a verdict about other generators, and it is not proof that a file is human-made. Record the actual check date or detector version, the raw result text, and the file’s transform history. The Gemini help cited here is the 2026-10-09 source for that product’s stated scope and limits.
DeepMind’s SynthID page, accessed 2026-10-09, describes the independent Detector as checking image, video, and audio from Google and participating partners, including OpenAI, NVIDIA, and Kakao. Apple is described as coming soon, not as already supported. That page does not claim universal partner, model, media, or version coverage. Uploading a file to that Detector is an external check, different from local container inspection.
Primary references
Why a missing credential proves less than people expect
Provenance is strongest when you have a positive, verifiable chain. The absence of that chain is ambiguous. The image may have come from software that never added C2PA, from a generator that uses a different watermark, from a social platform that transformed the file, or from an editor that exported a fresh container without carrying the original metadata forward.
This is why “AI detector” pages should not turn one missing field into a certainty. A useful checker should report evidence by layer: metadata, C2PA, watermark support, and visual signals. It should also make clear which systems each check can and cannot identify.
What to save if provenance matters
If you are a creator, publisher, investigator, or newsroom, keep the original file rather than only a social-media derivative. Preserve the file hash, the credential, and the verification result. If you need a privacy-safe public copy, create that copy separately rather than destroying the only provenance-rich original.
Questions about checking the file
When using a Content Credentials verify tool, what information can you find?
Issuer, actions such as created or edited, and the tools named on those actions, when the manifest is still in the file. Ingredients show up when they are listed. Social download counts do not. Width and height can sit in an EXIF assertion, or you read them from the file. A missing manifest is inconclusive.
How do I check C2PA Content Credentials in an image?
Open the original JPEG, PNG, or WebP in the Content Credentials checker, then choose image. Read whether a manifest is still there, who signed it, and which actions it lists. The C2PA viewer online is the same inspector. Optional Verify with OpenAI is a separate click and sends the file. This page is the how-to. The definition of the standard stays on what C2PA is.
Can this check remove the credential or an AI info label?
No. Inspection reads the packet. Metadata Cleaner can drop a supported embedded packet from a sharing copy. A pixel watermark such as SynthID remains. An Instagram, Facebook, TikTok, or YouTube label can remain. See Instagram AI info.
Inspect before you trust or strip
Local inspection stays in the browser after Choose image. Optional OpenAI verification sends the selected file. Keep the original if the credential matters.
Open the Content Credentials checker →