SynthID Detector

Guide · Provenance standards

C2PA vs SynthID: what is the difference?

C2PA Content Credentials and SynthID are often mentioned together because Google and OpenAI use both. They solve different problems. Content Credentials are a signed record of where a file came from and how it changed; SynthID is a watermark hidden in the content itself. Knowing which is which explains why an image can lose one and keep the other, and which tool can read each.

Check an image

In short

  • C2PA is an open, royalty-free standard for signed provenance records that travel with a file.
  • SynthID is Google DeepMind's invisible watermark embedded in the pixels, audio or text.
  • Content Credentials say much more but are stripped by screenshots; SynthID says less but is built to survive edits.
  • Google and OpenAI attach both to their AI images so one can back up the other.

What C2PA Content Credentials are

The Coalition for Content Provenance and Authenticity (C2PA) publishes the technical standard behind Content Credentials. According to the C2PA, a Content Credential, also called a C2PA manifest, is a tamper-evident, cryptographically signed data structure that travels with an image, video, audio file or document. It can record how the content was created, which tools were used, when it was made and how it has changed.

Each manifest is signed by a product whose certificate comes from a certification authority on the C2PA Trust List, and contains hashes of the asset and of the provenance data, so any modification breaks the link. The standard is open and royalty-free, works for camera photos as well as AI output, and is not a form of DRM: it records history rather than restricting use.

What SynthID is

SynthID is a watermarking technology from Google DeepMind. For images and video it changes pixel values by amounts too small to see, spread across the whole frame; for audio it embeds an inaudible signal in the waveform. A matching detector reads the signal back. Google says it has watermarked over 100 billion images and videos and 60,000 years of audio since introducing SynthID, and that OpenAI, Kakao and ElevenLabs are bringing it to their AI content.

SynthID carries a simple message, that the content was generated or edited by a participating company's AI, rather than a full edit history.

Side by side

The two systems differ on almost every practical point.

  • Where it lives: C2PA in a signed block stored with the file; SynthID inside the pixels or audio.
  • What it says: C2PA names the signer, the tool, the AI source type and the edit history; SynthID says the content came from a participating company's AI.
  • Screenshots and re-uploads: C2PA is usually removed; SynthID is designed to survive many such changes.
  • Who can read it: anyone with open C2PA tools for Content Credentials; only the operator's own detector for SynthID, such as the Gemini app for Google AI or openai.com/verify for OpenAI.
  • Scope: C2PA covers AI and non-AI media, including camera captures; SynthID marks AI-generated or AI-edited content.

How they work together

OpenAI explains why it uses both: C2PA helps content carry detailed context, while SynthID helps preserve a signal when metadata does not survive. Google's images do the same, and their Content Credentials state that a SynthID watermark was applied. The C2PA itself describes watermarks and fingerprints as 'soft bindings' that can help rediscover a Content Credential after it has been stripped, an approach it calls Durable Content Credentials.

In practice, a file fresh from Gemini or ChatGPT carries both. After a screenshot or a trip through social media, the credentials are usually gone and the watermark is what remains.

How to check each one

For Content Credentials, upload the original file to a C2PA reader, such as the free check on this site, which validates the signature against the C2PA trust list and shows whether the credentials declare a SynthID watermark. For the SynthID watermark itself, use the operator's checker: the Gemini app for Google AI content and openai.com/verify for OpenAI images. Google also announced in May 2026 that it is adding C2PA verification to the Gemini app and bringing SynthID verification to Search and Chrome.

Limits of both

Neither system can label content from tools that do not use it, and many open-source image models attach no provenance at all. A missing manifest or a negative watermark check therefore does not prove that an image is real. Content Credentials can show that a photo came from a camera, as Google's Pixel 10 does for photos taken with its camera app, which is one of the few positive signals for authentic media.

Frequently asked questions

Is SynthID part of C2PA?

No. SynthID is Google DeepMind's watermark; C2PA is an open standard for signed provenance records. They can be linked: Google's Content Credentials state that a SynthID watermark was applied, and C2PA allows watermarks as soft bindings.

Which survives a screenshot?

SynthID is designed to survive screenshots and many edits. Content Credentials are usually removed by screenshots and most social platforms.

Can anyone read SynthID?

No. Google's Gemini app reads Google's SynthID, and openai.com/verify reads OpenAI's. Content Credentials, by contrast, can be read and validated with open C2PA tools.

Do camera photos have C2PA?

Some do. Google says Pixel 10 was the first smartphone to provide Content Credentials for images in its native camera app, and Content Credentials are used for non-AI media by camera makers and news organisations.

Sources

Updated 2026-10-29. SynthID Detector is an independent service and is not affiliated with Google, Google DeepMind or OpenAI. SynthID is a Google technology. Results describe the provenance data found in a file and are not legal proof of how an image was made.