What happened
Google has opened its SynthID Detector to the public. It is a website where anyone can check whether an image, video or audio file was made with AI. It reads invisible watermarks inside JPG, PNG, MP4 and MP3 files.
The watermarks come from models like Gemini, Veo and Lyria. They also come from partners: OpenAI, Nvidia and Kakao use SynthID for media, and Google says Apple is expected to join soon. Google claims more than 180 billion images and videos now carry the mark.
The same detection is already built into Google Search, the Gemini app and Chrome. Google says that adds up to one million verification requests a day.
There is a login wall. Google, OpenAI and Apple accounts all work. Each user gets a daily quota of about 10 image, video and audio checks.
Google engineers told Ars Technica the cap exists to stop people from probing the checker to build watermark-removal tools. The fuzzy "approximately" is likely because near-identical uploads can trigger a lockout sooner.
Why it matters
Here is the catch, and it is a big one: SynthID is not a universal AI detector. It flags only content carrying a SynthID watermark. A clean result means "no SynthID found", not "a human made this".
Other detection systems share the same limit, usually verifying only their own provider's marks. Plenty of AI content comes from tools that use different watermarking, or none at all. OpenAI built its own separate system for text.
Still, one public place to check the big names is a real step. Before, provenance checking was scattered across vendors. Now a skeptical aunt with a suspicious photo has a single destination.
A clean result means no SynthID was found, not that a human made the thing.
What it means for Europe
Europe is the likely audience for this kind of tool. Transparency and content provenance rules in the EU lean on exactly this idea: machine-readable marks that say where media came from. A public detector from a major provider shows what that looks like at scale.
It also shows the limits regulators will wrestle with. A watermark check covers participating providers only, and the daily cap means it is a spot-check tool, not a newsroom-grade pipeline. Expect this to feature in EU debates about what "good enough" labeling means.
What to do with it
- Check the weird stuff.Run the suspicious viral image or clip through the detector before you share it.
- Read negatives carefully.No watermark found only means no SynthID, so keep your normal skepticism switched on.
- Budget your quota.About 10 checks a day and a login are required, so save them for files that matter.
- Look for the built-in version.Search, the Gemini app and Chrome already carry the detection, so you may not need the site at all.
Ciao, sapiens. Trust, but verify, and then verify the verifier.


