Google just made its AI watermark detector free for everyone

Telling whether a viral clip or audio recording is real or AI-generated has become a daily headache for anyone on the internet.

Google is now rolling out its in-house lie detector to the public, launching a dedicated SynthID Detector website that anyone can use for free.

First introduced in 2023, SynthID embeds imperceptible watermarks directly into AI-generated media. While Google gave early access to select journalists, media pros, and researchers at Google I/O last year, the new web tool is now available globally in English.

A massive footprint and a growing alliance

The scale Google is operating at here is staggering. Since 2023, the company has watermarked over 180 billion images and videos, alongside roughly 240,000 years of audio content.

Every piece of media generated by Google's own roster, including Nano Banana, Veo, and Lyria models, as well as products like Gemini, Flow, ProducerAI, and Vids, gets tagged with SynthID by default.

Google is not keeping this garden walled off. The detector can also identify AI content produced by partners like OpenAI, NVIDIA, and Kakao, with Apple scheduled to join the lineup soon. OpenAI also operates its own verification site.

The tool handles virtually every everyday format you can throw at it. That includes image files like JPG, PNG, WEBP, AVIF, HEIC, TIFF, and GIF; MP4, MOV, and WEBM for video; and audio tracks across WAV, MP3, OGG, FLAC, AAC, and M4A.

Google already baked SynthID checks directly into Google Search, the Gemini app, and Chrome, noting that these native integrations already field more than 1 million verification requests every single day.

Why it matters

For regular users and creators, having a friction-free, universal web portal to check suspicious clips or voice tracks is an immediate win. You don't need technical skills or specialized software to get a quick sanity check.

Strategically, Google is playing this very smartly. A watermarking standard is only as good as its network effects. By teaming up with heavyweights like OpenAI, NVIDIA, and soon Apple, Google is positioning SynthID as the default detection backbone for the entire industry.

This move puts clear pressure on Microsoft and Meta, both of which have been championing their own distinct watermarking and verification standards.

That said, we think users should be careful not to treat this as a bulletproof truth engine. As industry testing has repeatedly shown, watermarking detectors are not infallible and still occasionally fail to spot media made by their creators' own models. More critically, open-source models that strip out watermarks altogether remain completely outside this net. It is a big step forward for internet hygiene, but it won't wipe out synthetic deception overnight.