What are Content Credentials and how do you check them?
Content Credentials are a signed record attached to a photo, video or other file. They can show which device or app made it and how it was changed. This guide explains what they tell you, how to read them for free, and what to do when a file has none.
1. What Content Credentials are
Content Credentials follow an open technical standard from C2PA, the Coalition for Content Provenance and Authenticity. The Content Authenticity Initiative describes them as a "nutrition label" for digital content.
A credential can record:
- who or what produced the file, and when;
- which camera, app or AI tool was used;
- what edits were made afterwards;
- which earlier files went into it.
The record is cryptographically signed and travels with the file. If the file is changed after signing, the link breaks and checking tools can show that it was tampered with.
2. Where they come from
Credentials are only added by devices and apps built to do so.
- Phones. Google's Pixel 10 phones add them to every JPEG photo taken with Pixel Camera.
- Cameras. The Leica M11-P was the first camera with them built in, in 2023. Sony also makes cameras that can sign pictures in the C2PA format.
- AI tools. OpenAI says images made with ChatGPT and its API carry C2PA metadata.
- Platforms. LinkedIn shows a C2PA icon on images and videos that carry credentials. Select it to see the details.
3. How to check a file
You will need the file itself, as close to the original as you can get.
- Save the picture or video to your device. Avoid taking a screenshot, because a screenshot is a new file with no credentials from the original.
- Go to Content Credentials Verify, or Adobe's Inspect tool.
- Select the file, or drag it onto the page.
- Read the panels that appear. If the file has several stages, look through each one.
Do not upload private photos of yourself or other people to any online tool.
4. How to read the result
Look for these things:
- The signer. This is the organisation or product that signed the record. On Inspect, a signer whose certificate is not on the C2PA trust list is shown as "Unrecognized".
- The app or device. A camera name suggests a capture. An AI tool's name tells you AI was involved.
- The actions. These list what was done to the file, such as edits.
- The ingredients. These are the earlier files used to make this one. You can compare them with the final version.
A stated AI tool is strong evidence, because the record is signed by the tool's maker. A camera record is good evidence of where a picture came from. Neither one tells you whether the caption attached to it is honest.
5. When there are no credentials
Inspect will say "No Content Credential". This is the result you should expect for most files you find online.
Missing Content Credentials do not prove a file is fake. They only mean this check cannot help.
Credentials can be separated from a file by accident or on purpose. OpenAI warns that editing, converting or sharing a file can remove C2PA metadata. Plenty of real photos come from devices that never added any.
On Inspect you can also search for similar content in the Content Credentials Cloud, which may find a copy that kept its record. If nothing turns up, move on to other checks. Try a reverse image search, then the rest of our guide to checking if an image is AI-generated.
6. What credentials cannot tell you
The Content Authenticity Initiative says credentials are not meant to show whether content is "real". They describe where a file came from, and you judge the rest.
- They do not check the story. A genuine, signed photo can still be shared with a false caption. See misleading context.
- A name is not proof of identity. Creators do not have to include their identity, and a name shown may not match the content's owner.
- The history of each part may be incomplete. C2PA says that checking the full history of content built from several sources needs access to each source's own data.
- Devices can have flaws. In 2025 a user found a way to make a Nikon Z6 III sign an image that did not come from its sensor, including an obvious AI image. He got around the check without breaking the cryptography. Nikon suspended its service.
7. Invisible watermarks
Metadata can be lost, so some companies also hide a watermark inside the pixels or sound. The best known is SynthID, from Google DeepMind. OpenAI now adds it to images too. This is a separate system from Content Credentials, and it needs its own checker.
The SynthID Detector looks for watermarks from Google and partners including OpenAI, Nvidia and Kakao. OpenAI also has a checking tool at openai.com/verify, which covers its own images.
Treat these results with the same care. OpenAI says finding nothing does not mean a file is human-made. Full Fact found Google's detector flagged five images that the people who posted them said were genuine, and it advises using other checks as well.
8. Decide what you can say
- Verified. Valid credentials from a recognised signer, plus a reliable source, support what is being claimed about the file.
- AI-generated or altered. The credentials or a watermark check name an AI tool or show changes, and nothing else contradicts that.
- Misleading context. The credentials show a genuine capture, but it is being used to tell a false story.
- Unverified. There are no credentials, or they do not settle the question.
Most files will end up as unverified at this step. That only means you need other checks before you say more.
In the news
Recent reports from RELTY's AI news that show why this check matters.
- 4 November 2026 · Maldita.es
Video of a Spaniard denied benefits was made with AI (in Spanish)
Maldita.es cites a result from Google's SynthID Detector on the clip. Watermark checks work on video as well as still images.
- 29 October 2026 · Maldita.es
Netanyahu image naming Spain was altered with AI (in Spanish)
Maldita ran the image through OpenAI's verification tool, which found a SynthID watermark. The fact-checker then traced it to an older version without the word. A provenance signal works best alongside a search for the original.
Sources
The pages this guide relies on, checked on 1 November 2026.
- FAQs. Coalition for Content Provenance and Authenticity.
- How it works. Content Authenticity Initiative.
- Content Credentials Verify. Content Authenticity Initiative.
- Using the inspect tool on Adobe Content Authenticity (Beta). Content Authenticity Initiative open-source documentation.
- How Pixel and Android are bringing a new level of trust to your images with C2PA Content Credentials. Google, September 2025.
- Leica launches world's first camera with Content Credentials built-in. Adobe, October 2023.
- Camera Authenticity Solution. Sony.
- Content credentials. LinkedIn Help.
- Nikon struggles with security problems in photo authentication. heise online, September 2025.
- Content provenance. OpenAI developer documentation.
- SynthID. Google DeepMind.
- Google updates AI detector after Full Fact investigation finds it flagging apparently real images. Full Fact, October 2026.
How to tell if an image is AI-generated
Why AI detectors disagree, and how to read their results
How to tell if text was written by AI
How to reverse image search a photo
Browse the RELTY directory of tools
Links to other organisations are for reference only. RELTY has no partnership with, or endorsement from, any of them.
