Since 7 October, the SynthID Detector, Google’s tool for spotting content generated by artificial intelligence, is no longer reserved for professionals: it is open to everyone, worldwide. For now it works in English only, and Google has not said when other languages will follow.
What it actually does
SynthID is an invisible watermark. It is embedded in the file at the moment a model generates an image, a video or an audio track, and it can be neither seen nor heard. The Detector works in reverse: you upload a piece of content and the tool checks whether that watermark is present.
The launch comes from the Google DeepMind blog, signed by Pushmeet Kohli. Google says the technology dates back to 2023 and has so far watermarked more than 180 billion images and videos, plus 240,000 years of audio. A preliminary version of the Detector was given to information professionals last year; now access is widening to everyone.
The limit not to forget
The Detector only recognises what has been watermarked. If an image comes from a generator that does not use SynthID, the result says nothing about its origin: “watermark not found” does not mean “authentic content”. 9to5Google makes the same point, stressing that the tool only works when whoever generates the content embeds the watermark upstream.
Who is on board
Besides Google’s own models, Mountain View names OpenAI, NVIDIA and Kakao as partners. On Apple the wording is more cautious: its arrival is described as imminent (“soon”), with no date. For iPhone users this matters, because images created with Apple’s tools cannot yet be identified this way.
The Detector sits alongside the checks Google has already built into Search, the Gemini app and Chrome. According to the company, these features now routinely handle more than a million requests a day.
Who it is for
- People who work with images: journalists, designers and social media managers can check a photo or video in seconds before sharing it.
- People who receive suspicious content: a quick check on a viral image in a chat or on social media helps, if the file comes from a partner generator.
- People who create with AI: knowing that their material is watermarked makes it easier to prove where it came from.
What we think
Opening the tool to everyone is a useful step, but its strength depends on participation. The more generators adopt SynthID, the more meaningful the Detector’s result becomes; as long as many models, especially open-source ones, stay outside, a negative result should be read with caution. Apple’s arrival would be the most important signal, because it would put the watermark on the smartphones of millions of people.
Then there is the matter of language: an English-only interface holds back use by the Italian public. The practical advice is to treat the Detector as one more clue, to be paired with checking the source and the context, not as a verdict. To find out more about Google’s other AI tools, also read Gemini 4 Argon and limited access.
Written by the SpazioiTech newsroom with the help of AI tools, from facts verified across at least two independent outlets, and checked before publication. Spotted a mistake? Tell us.
Translated from the Italian original. Read in Italian
