isGenAI.com

Glossary

SynthID: Google’s invisible watermark for AI content

SynthID is Google DeepMind’s technology for embedding an imperceptible watermark into AI-generated images, audio, video and text at generation time, designed to survive common edits and to be detected by Google’s own tools rather than read from metadata.

Definitions and isGenAI behaviour as of October 2026. Updated .

See the term in a real reportFrequently asked questions
Code-drawn illustration for SynthID

Illustration · PNG file

  1. Signature + IHDR
  2. iTXt XML:com.adobe.xmprecord
  3. tEXt parameters
  4. tEXt prompt / workflow
  5. tEXt invokeai_metadata
  6. eXIf
  7. caBX C2PA manifest
  8. IDAT pixels
  9. IEND

AI source type declared

The file labels itself as AI-made.

IPTC DigitalSourceType such as trainedAlgorithmicMedia is read from XMP and shown with its exact value.

Check an image →
AI source type declared: where this record sits inside a PNG file and what the isGenAI report says about it. The picture is a code-drawn illustration.

What SynthID is

SynthID changes pixel values in a way people cannot see but a matching detector can recognise, and Google states it is built to withstand modifications such as cropping, compression and filters. Google applies it to Gemini and Imagen images and offers a check inside the Gemini app. OpenAI’s help centre states that images generated with ChatGPT, Codex and the OpenAI API also include a SynthID watermark alongside C2PA metadata.

SynthID is a soft binding: it lives in the content, not in a metadata container, so stripping metadata does not remove it.

How isGenAI relates to SynthID

isGenAI does not detect SynthID. The checker reads file records, C2PA manifests, IPTC source types, AI statements and saved prompts, and matches copies by hash and fingerprint. When a Gemini or ChatGPT image has been stripped of its manifest, isGenAI reports that no file evidence was found and points to the vendor’s own watermark check.

Reporting only what the file records is deliberate: it keeps every finding retraceable.

Where to check for SynthID

Google’s Gemini app can be asked whether an uploaded image was made with Google AI. Google has stated plans to extend verification to C2PA credentials. For other vendors’ watermarks, use that vendor’s verifier.

Because SynthID and C2PA answer different questions, a vendor that uses both gives checkers two independent routes: the signed file record for anyone with a C2PA validator, and the watermark for the vendor’s own detector after the metadata has gone.

How to use this term when checking a file

  1. Upload the original file to the AI image detector or the tool named on the page.
  2. Find the finding that uses this term under “How we checked” and read the source field it names.
  3. Weigh it by kind: a signed record, a declared field, a saved prompt and a matching copy are different strengths of evidence.
  4. Remember that absence of a record is not evidence of a human author.

See the term in a real report

Frequently asked questions

Can isGenAI detect SynthID?

No. isGenAI reads file records and matching copies. Use Google’s Gemini app to check for SynthID.

Does SynthID survive a screenshot?

Google states it is designed to survive common modifications. Verify with Google’s check; isGenAI makes no claim about watermarks.

Which images carry SynthID?

Google documents it for Gemini and Imagen images, and OpenAI documents it for images from ChatGPT, Codex and its API as of October 2026.

Is SynthID the same as C2PA?

No. SynthID is an invisible pixel watermark; C2PA is signed metadata. They complement each other.

Sources

Definitions follow the published standards linked above; the isGenAI behaviour described is the deployed checker as of October 2026.