Google's new AI detector can say yes, but its no means little
Medium: available today, and you can use it any time you doubt a photo, video or audio clip.
What happened
Google opened its SynthID Detector to everyone on October 7. It's a website, synthid.com, where you can check an image, video or audio file for a hidden marker left by Google's AI tools or those of partners OpenAI, NVIDIA and Kakao. Apple support is coming "soon," Google says. It's free but requires a Google, OpenAI or Apple login, Engadget reports. It works worldwide, in English only for now.
What's actually going on
When an AI tool that uses SynthID makes a picture, video or sound clip, it hides a pattern inside the file itself. That hidden pattern is called a watermark. For a picture, it means nudging tiny details in ways your eyes can't see. For audio, it's changes your ears can't hear. Google says it has marked over 180 billion images and videos and 240,000 years of audio this way.
The detector looks for that pattern. If it finds it, that's a good sign the file came from Google's tools or a partner's. What's new is who can ask. Before, checking was built into Google Search, Chrome and the Gemini app (Google's AI assistant), which Google says now handle over 1 million requests a day. An early version of this website was only for journalists and other media workers. Now anyone can use it.
This design explains the weak spots. The detector can only find marks that were put there. A fake from a company that doesn't use SynthID has no mark, so it comes back clean. Engadget notes companies use different marking systems, so no single checker covers everything. Meta's own tool, for example, only spots images from its newest models.
Marks can also be stripped in some situations, Engadget says. And the tool can't tell you whether AI made the whole thing or just touched it up.
The word you'll hear
Watermark
An invisible pattern an AI tool hides inside a picture, video or sound file it makes, so a checker can spot it later.
In your life
Next time a shocking photo or a clip of a famous voice hits your group chat, you can check it in roughly a minute. A "yes" means something. A "no" means much less.
Be careful about
A clean result doesn't mean it's real. It only spots marks from Google and its partners, marks can sometimes be stripped, and it won't say if something was fully made by AI or just edited.
Where to get it
Free at synthid.com with a Google, OpenAI or Apple account. The sources mention no paid version. Try it first on a photo you took yourself, so you know what a "no" looks like.
Go deeper
We're making it easier to identify AI-generated content globally.
Google Blog, October 7, 2026
Also checked against: Engadget
How was this brief?
Get one of these every morning
One story, about five minutes, in your inbox by 7 AM Eastern. No spam, and unsubscribing takes one tap.