SynthID - The Invisible Safeguard
In the era of Agents, Google has unveiled an array of innovations starting from Gemini 3.5 Flash processing tokens at breakneck speeds, Gemini Omni generating "anything from any input," and even screenless Audio Glasses designed in tandem with Samsung
But with AI becoming an invisible fabric of our daily lives—autonomously drafting our workflows and generating hyper-realistic video—a massive, underlying problem takes center stage: authenticity.
Research shows that people can correctly identify high-quality deepfake videos only about a quarter of the time. As AI-generated content becomes completely indistinguishable from reality, how do we protect the digital ecosystem?
Google’s answer is a massive, multi-partner expansion of SynthID—and it might just be the most important announcement of the entire keynote
What is SynthID?
If you haven't been tracking it, SynthID is Google DeepMind’s digital watermarking technology. Unlike traditional watermarks that overlay a visible logo onto an image, SynthID embeds a digital signature directly into the data pixels, audio waves, or text tokens.
SynthID is completely invisible to the human eye and ear, but instantly readable by a dedicated detector. Crucially, it’s highly resilient—it survives editing, cropping, compression, and screenshotting.
Since its quiet launch a few years ago, the scale of this project has exploded. Google revealed that SynthID has now watermarked:
Over 100 billion images and videos
More than 60,000 years of audio assets
How SynthID works?
SynthID embeds digital watermarks directly into AI-generated images, audio, text or video. The watermarks are embedded across Google’s generative AI consumer products, and are imperceptible to humans – but can be detected by SynthID's technology.
AI-generated image and video
SynthID adds an invisible digital watermark to an AI-generated image/video segment. The watermark doesn’t change the image or video quality. It’s added the moment content is created, and designed to stand up to modifications like cropping, adding filters, changing frame rates, or lossy compression.
AI-generated audio
SynthID embeds a watermark into any audio generated or published through the AI music generation model Lyria or the podcast generation feature of Notebook LM. It’s inaudible to the human ear, and can’t be altered by common modifications like adding noise, MP3 compression, or changing the speed of the track.
AI-generated Text
Large language models generate text one word (token) at a time. Each word is assigned a probability score, based on how likely it is to be generated next. So for a sentence like “My favourite tropical fruits are mango and…”, the word “apples” would have a higher probability score than the word “ships”. SynthID adjusts these probability scores to generate a watermark. It's not noticeable to the human eye, and doesn’t affect the quality of the output.
SynthID’s Cross-Industry Alliance
Until now, SynthID was primarily a Google-centric sandbox. But a watermark tool is only as strong as its adoption rate. If only Google uses it, the rest of the internet remains a wild west of unverified media.
At I/O ‘26, Google dropped a bombshell compliance alliance. While Nvidia signed on last year, three new absolute titans are officially adopting SynthID:
OpenAI
ElevenLabs
Kakao
By uniting the creators of ChatGPT/Sora (OpenAI) and the leading AI voice-generation platform (ElevenLabs), SynthID is rapidly becoming the universal, cross-industry protocol for AI transparency.
Detecting SynthID watermarks in Gemini Want to check if an image, video or audio clip was generated, or edited, by Google AI?
Simply upload the image, video or audio clip to your Gemini chat, and ask if it’s been created or altered by Google AI. Gemini will check for a SynthID watermark, and let you know if it finds one.
Moving Into the Browser: Integration with Search and Chrome
It’s one thing for backend systems to watermark content; it’s another for everyday users to actually see it. Google is bringing this verification right to the front lines by expanding both SynthID and C2PA Content Credentials into Google Search and Chrome.
When you browse the web via Chrome or look at an image in Google Search, the browser will actively parse the content's metadata and invisible pixels. It will cleanly tell you:
The Origin: Whether the media came from a physical camera or an AI engine.
The Edit History: If a real photo was structurally manipulated using generative AI tools.
As we move aggressively into a world of autonomous AI agents, Google acknowledges that power requires accountability.
By building SynthID directly into Chrome and Search—and convincing its competitors like OpenAI to hop on board—Google isn't just shipping cool AI features anymore. They are actively building the safety railings for the next decade of the internet.
