Free · runs in your browser
Open a video or photo to read its C2PA manifest in plain English: who made it, with what, what was done to it, and whether AI was involved.
To see a video’s Content Credentials, open it here and go to Provenance. It tells you whether the file carries a C2PA manifest and, if it does, shows the app or device that wrote it (the claim generator), the actions recorded such as created or edited, any AI-generation assertion, the ingredients it was built from, and the signer’s name taken from the certificate.
This viewer reads the manifest; it does not check the cryptographic signature. To confirm a signature is valid and the file has not changed since it was signed, use the free verifier at contentcredentials.org/verify. Reading first is still useful: it works on files you would rather not upload anywhere, and it explains each entry.
The file stays on your device. Provenance also lists encoder and software tags and other signs that a file has been re-saved, which helps when there is no manifest at all. If your real question is whether a clip was made by AI, read how to tell if a video is AI as well: Content Credentials are one check among several.
Press Open a file (Ctrl/Cmd+O), drop the file on “Drop a video or photo here”, or paste it (Ctrl/Cmd+V). Videos and photos both work.
Choose Provenance in the section list (or press J/K to step through). The first line says whether Content Credentials are present.
Check the claim generator (the app or camera that wrote the manifest), the signer’s name from the certificate, and any AI-generation assertion (its digitalSourceType). The ⓘ next to each field explains what it means.
Look through the actions (created, opened, edited, converted) and the ingredients, which are the earlier files this one was made from, each with its own manifest when it had one.
To confirm the signature is valid and the file is unchanged since signing, open contentcredentials.org/verify in your browser and check the same file there.
Content Credentials are the public name for the C2PA standard, written by the Coalition for Content Provenance and Authenticity. An app or camera that supports it writes a manifest into the file: a set of statements (assertions) about how the content was made, bundled into a claim and signed with a certificate. The claim also includes a hash of the file’s media, so any later change to the picture or sound breaks the match.
| Entry | What it means | Example |
|---|---|---|
| Present / not present | Whether the file contains a manifest store at all | Most videos you download today have none |
| Claim generator | The software or device that wrote the latest manifest | An editing app and its version, an AI video tool, a camera’s firmware |
| Signer | The organisation named in the signing certificate | The vendor of the tool or camera, or a news organisation |
| Actions | What was done, in order | c2pa.created, c2pa.opened, c2pa.edited, c2pa.converted, c2pa.placed |
| AI-generation assertion | The digitalSourceType: how the content came to exist | “Created by a generative AI model”, “Composite with AI elements”, “Digital capture” |
| Ingredients | Earlier files this one was built from | The original camera clip, a stock image, an AI-generated background |
C2PA borrows its wording for how media came to exist from the IPTC digital source type vocabulary. The raw value is a web address ending in a short term, and the ⓘ next to the field explains it.
| digitalSourceType | Plain English | What it usually tells you |
|---|---|---|
trainedAlgorithmicMedia | Made by a generative AI model | The whole picture came from an AI tool, typically from a text or image prompt |
compositeWithTrainedAlgorithmicMedia | Mixed with AI-generated parts | Real footage or images combined with AI elements, such as a generated background or an extended frame |
algorithmicMedia | Made by an algorithm without training data | Procedural graphics, not generative AI |
digitalCapture | Captured by a camera | The signer says it came from a sensor. Only as trustworthy as the signer |
compositeSynthetic | A composite of several sources | Edited from multiple inputs, with or without AI |
An assertion is a claim made by whoever signed it. A valid signature from a well-known AI vendor saying trainedAlgorithmicMedia is strong evidence the clip came from that tool. An unsigned or unverifiable manifest saying digitalCapture is not evidence of anything until the signature checks out.
This viewer does not verify signatures. It shows the signer’s name as written in the certificate, and a forged or broken manifest can name anyone. For a validity check, use contentcredentials.org/verify.
A verifier answers three questions that reading alone cannot:
Other places can check too. Google’s Gemini app verifies Content Credentials on images, video and audio (version 2.2 and later, from products on the C2PA conforming products list), and can look for Google’s SynthID watermark in videos under 100 MB and 90 seconds. Google Photos on Android shows a “How this was made” summary under More → About when a file has credentials.
Support is growing, but it is patchy, and stills are ahead of video. The table lists what each vendor says on its own pages, checked on 7 November 2026. Claims change quickly, so treat it as a snapshot.
| Source | What the vendor states | Kind |
|---|---|---|
| Adobe Firefly | Firefly Video Model outputs include Content Credentials | AI generator |
| Microsoft Bing Video Creator | C2PA content credentials on every video it generates | AI generator |
| Google (Gemini app, Flow, YouTube) | Content made or edited with the Omni model carries SynthID and C2PA; all Flow outputs carry SynthID | AI generator |
| OpenAI Sora | Sora videos carried C2PA metadata; OpenAI announced in March 2026 that it is discontinuing Sora | AI generator |
| Sony cameras | Camera Verify for video since October 2025 on models including the a1 II, a9 III, FX3, FX30 and PXW-Z300, with a signing licence | Camera |
| Google Pixel | Signs photos in the Pixel Camera app; video support announced in May 2026, while the camera help page still said videos are not supported | Phone camera |
| TikTok | Reads Content Credentials to label AI content uploaded from other tools, and has said it will attach credentials to TikTok content | Platform |
| Shows the C2PA icon on signed image and video posts; tap it to see the credentials | Platform |
“No Content Credentials” is the most common result by far, and on its own it means very little. Most cameras and phones do not sign video yet. Many editing apps drop manifests they do not understand when they export. Social platforms re-encode uploads and usually serve a new file without the original manifest. Screen recordings start a new file from scratch.
For a full judgement on a suspicious clip, combine provenance with the other checks in the AI video guide. For signs of re-saving and editing beyond C2PA, see video forensics. To see every other field the file holds, the full file info report lists them all.
No. It reads and explains the manifest, including the signer’s name from the certificate, but does not check the signature or the trust list. Use contentcredentials.org/verify, or the Gemini app, to confirm a signature is valid.
Most videos have none: few cameras sign video, many apps drop manifests on export, and platforms re-encode uploads. Absence tells you nothing about whether the clip is genuine.
Often, yes. Platforms usually re-encode videos and serve a new file, and unless the platform writes its own manifest, the original one is gone. A downloaded copy of a post is rarely the file that was uploaded.
LinkedIn says it shows the C2PA icon on image and video content signed with Content Credentials, and tapping it shows the credential details. It means a manifest was found, not that the content is true or untrue.
Anyone can write a manifest, but they cannot sign it as Adobe or Sony without that company’s private key. That is why verification matters: a forged manifest fails the signature or trust check, while this viewer would still display what it says.
No. The manifest is read in your browser and nothing about the file is sent anywhere. If you then use contentcredentials.org/verify, that is a separate site with its own privacy terms.
No. A manifest is metadata; watermarks such as Google’s SynthID are woven into the pixels and audio and survive a metadata strip. This site cannot read or remove invisible watermarks.