News · Oct 7, 2026 · updated Oct 9
SynthID Detector, explained honestly
On October 7, 2026 Google opened SynthID Detector globally in English so anyone can check whether image, video, or audio was watermarked by Google or its partners. Here's what the tool actually does — and what it doesn't.
What just shipped
Google's DeepMind announcement and the official @Google post say the verification portal is now open worldwide in English. You upload media and ask: was this made (or edited) with a SynthID-watermarked AI system?
- Portal: synthid.com
- Detects media from: Google, OpenAI, NVIDIA, Kakao — soon Apple
- Formats: images, video, and audio
- Login: Google, OpenAI, or Apple account
- Quota: about 10 checks per day per user (per reporting on Google's rollout)
What SynthID is (and isn't)
SynthID is Google DeepMind's imperceptible digital watermark for AI-generated media. It embeds a signal in the pixels / frames / audio that humans usually cannot see or hear, but a matching detector can read. It is not the same thing as:
- Claude's statistical text watermark (woven into word choice)
- C2PA Content Credentials (signed file metadata)
- Zero-width / invisible Unicode characters that sneak into copied text
Those are different layers. Confusing them is how people end up buying fake "remove AI watermark" tools.
How to check a file with SynthID Detector
- Open synthid.com.
- Sign in with a Google, OpenAI, or Apple account.
- Upload an image, video, or audio file.
- Read whether a SynthID watermark was detected for a supported partner.
The public portal is meant for a clear yes/no-style watermark check. It will not magically label every AI file on the internet.
Honest limits
- Partner coverage only. No SynthID mark from Claude text, Grok, or many open-weight / regional models means the portal will not flag them.
- Absence ≠ human. No watermark found does not prove a human made the file. Screenshots, heavy re-encodes, or unsupported tools leave no SynthID signal.
- Presence ≠ full authorship story. A mark means a supported AI system likely generated or edited the media; it does not by itself settle every legal or academic dispute.
- Login + daily quota. You need a Big Tech account, and checks are rate-limited.
Search terms this page targets
People hitting Google's news wave usually search variants like:
SynthID Detector · Google SynthID Detector · SynthID watermark · check if image was made with AI · AI generated media detector · SynthID OpenAI NVIDIA Kakao Apple · synthid.com · identify AI generated content Google
Where NoAtMark fits
We do not strip SynthID, and we will not ship a "bypass detector" page. That fights transparency and contradicts our red lines.
What we do own is the text hygiene layer after you copy from ChatGPT, Claude, Gemini, or Docs:
- Find and clean zero-width and hidden Unicode
- Scan paste for injection / hidden text before publishing
- Explain Claude text marks and C2PA honestly
Paste AI text → see the invisible characters
Media watermarks go to SynthID Detector. Copied text still needs a hygiene pass before it hits CMS, RAG, or code.
🔍 Open the free scanner C2PA checker