All posts

Blog

Google opens SynthID Detector to everyone, with checks for OpenAI, NVIDIA and Kakao watermarks

FindAmNow
GoogleSynthIDAI watermarkingdeepfakestrust and safety

On 7 October 2026 Google made SynthID Detector available globally in English. Anyone who signs in can check an image, video or audio file for SynthID watermarks from Google and partners including OpenAI, NVIDIA and Kakao. It is not a general AI detector, and unwatermarked AI content will not show up.

On 7 October 2026 Google made its SynthID Detector available to anyone, globally, in English. The tool checks whether an image, video or audio file carries a SynthID watermark, the invisible signal that Google and a growing list of partners embed in content their AI systems generate.

The announcement came in a Google blog post by Pushmeet Kohli, VP of Science and Strategic Initiatives at Google DeepMind.

What changed

Google introduced an early version of the detector in 2025 for journalists and media professionals. The new version opens it to the public and widens what it can recognise.

According to Google, the detector can now tell you whether a file was made with AI from:

  • Google
  • OpenAI
  • NVIDIA
  • Kakao
  • Apple ("soon", per Google)

Ars Technica reports the previous detector only checked for Google's own version of SynthID, so content from partners such as ChatGPT images would not have been flagged there even if it carried a watermark. The new site is meant to check all partner watermarks in one place.

How it works in practice

  • Sign-in required. Ars Technica and The Next Web report you can sign in with a Google, OpenAI or Apple account.
  • Daily quota. Ars Technica quotes Google as giving each user "a daily usage quota of approximately 10 image, video, and audio checks."
  • Simpler answers. Per Ars Technica, the public site reports whether SynthID was used to produce or edit the media, rather than highlighting the specific watermarked regions the earlier internal tool showed.
  • Made vs. edited. MacRumors notes the tool cannot distinguish AI-created media from AI-edited media.

Google says the same verification is already built into Search, the Gemini app and Chrome, and that those built-in checks now handle more than 1 million requests a day. The company also says it has watermarked more than 180 billion images and videos, plus 240,000 years of audio, since SynthID launched in 2023.

The limits matter

This is not a general-purpose AI detector. It only finds SynthID watermarks. A negative result means no SynthID watermark was found. It does not prove a file is authentic:

  • Content from AI tools that don't use SynthID will not be flagged.
  • The Next Web notes watermarks can be weakened or removed in some cases.

Read a "no watermark" result as "unknown", not "real".

Why it matters

Fake images, cloned voices and synthetic video are now cheap to produce. A free checker that covers several major AI providers gives ordinary people, newsrooms and businesses a quick first test before sharing or acting on suspicious media.

For small businesses and online sellers, that can mean a quick check on a product photo, a "proof of payment" screenshot or a voice note that seems off. It should be one check among several, not the final word.

What to watch

  • Whether Apple's support arrives on schedule.
  • Whether more AI companies adopt SynthID, or whether a rival standard, such as C2PA content credentials, ends up covering more content.
  • Whether the roughly 10-checks-a-day quota changes for heavier users like newsrooms.

Sources

Source