Google Opens Its AI Watermark Checker Globally, With Support Beyond Gemini

Checking an image, video or audio file for a hidden AI watermark no longer requires asking Gemini. Google opened its SynthID detector globally in English on October 7, 2026, with a dedicated website at SynthID.com. The service checks signals from Google and supported partners—not every possible source of AI-generated media.

A hidden signal, not a judgment of appearance

SynthID works by embedding an invisible signal in the pixels of generated images and videos, or in an audio file’s waveform. The detector looks for that deliberately added watermark. Its result tells users whether a supported SynthID signal was detected in media produced or edited with AI, rather than deciding whether the content merely looks artificial.

The rollout also broadens which signals the checker recognizes. Google names OpenAI, Nvidia and Kakao as supported partners, with Apple support coming later. Previously, Google’s detector checked only Google’s version of SynthID. That meant a ChatGPT image carrying a partner watermark would not be flagged by Google’s checker, according to Ars Technica.

Google introduced an initial detector for media professionals last year. Outside its trusted-tester program, users previously had to ask Gemini to check for SynthID. The new public site returns a detection message; it does not provide the internal tool’s view highlighting image regions where the watermark appears.

More than 180 billionImages and videos watermarked

Google says SynthID has been added to more than 180 billion images and videos.

240,000 yearsAudio watermarked

Google also reports watermarking 240,000 years of audio content.

Public access, with an anti-abuse ceiling

The website requires sign-in and accepts Google, OpenAI and Apple accounts. Apple account access is separate from the forthcoming ability to detect Apple’s watermarks. Google says each user gets approximately 10 checks per day across images, video and audio—not a separate allowance of 10 for each format.

That ceiling has a security purpose. Google engineers told Ars Technica that limiting checks helps prevent people from using the detector to develop watermark-removal tools. Opening verification to everyone therefore does not mean offering unrestricted access: the same checking service could be used to test attempts to strip its signal.

No watermark is not an all-clear

A negative result does not establish that a file is authentic or human-made. Some AI providers use different watermarking technology; others use none. Media generated with those systems falls outside this checker’s coverage. The website will include a disclaimer explaining that AI content without a supported watermark may not be identified.

The dedicated site joins Google’s existing verification features in Search, the Gemini app and Chrome. Google says those features now regularly process more than one million requests a day. That figure describes the existing verification services, not usage of the newly opened website.