The tool was previously limited to journalists, media professionals, and researchers who joined a waitlist after a pilot version launched in May 2025.
The detector is built on SynthID, Google's invisible watermarking technology introduced in 2023. Google embeds imperceptible watermarks into everything its generative models produce, and the new website can spot those signatures on uploaded files.
The announcement was made on The Keyword by Pushmeet Kohli, Google DeepMind's vice president of science and strategic initiatives.
What the detector can (and can't) check
The website accepts a wide range of formats: images in JPG, PNG, WEBP, AVIF, HEIC, TIFF, and GIF; videos in MP4, MOV, and WEBM; and audio in WAV, MP3, OGG, FLAC, AAC, and M4A.
It flags watermarks from Google's own generation tools and models, including Gemini, Imagen, Veo, Lyria, Nano Banana, Flow, ProducerAI, and Vids.
Google says the detector also recognizes SynthID watermarks from partner companies OpenAI, Nvidia, and Kakao. Apple is reportedly planning to add support for the standard soon. OpenAI maintains its own verification page for checking SynthID content.
But the company is careful to note that this is not a general-purpose AI detector. The website's FAQ states it can only identify media from companies that have adopted SynthID technology. Anything produced by a model that does not watermark its output will pass through undetected.
180 billion watermarks and counting
According to Google, more than 180 billion images and videos now carry SynthID watermarks, along with the equivalent of 240,000 years of audio content. The company says people make about 1 million verification requests every day, a volume handled through verification built into Google Search, the Gemini app, and Chrome.
The public launch ends a period of restricted access. A select group of journalists and researchers tested the tool after it was previewed at Google I/O last year. Now the same capability is available to anyone who wants to verify a suspicious image in a group chat, a questionable video online, or an AI-generated audio clip.
Why it matters — and where it falls short
As AI-generated media becomes harder to tell apart from the real thing, verification tools are becoming part of the internet's trust infrastructure. Invisible watermarks give platforms and users a way to confirm provenance without degrading the quality of the content itself.
There are limits. Google requires users to sign in, and reports suggest each user is limited to roughly 10 checks per day, a quota the company says is designed to make it harder for attackers to develop watermark-removal techniques.
Microsoft and Meta maintain their own separate watermarking standards, and Google acknowledges that detection tools — including its own — are not infallible and can miss content from models outside the system.
By Hamza Faisal
