A photo, video or audio clip can now be checked for Google's AI watermark without requiring access to specialised verification tools.

Google has opened its SynthID Detector to the public, allowing anyone to upload an image, video or audio file and check whether it contains a SynthID watermark associated with AI-generated or AI-edited content.

The public launch marks a significant expansion of a tool Google first began testing with journalists, media professionals and researchers.

The timing is important. As generative AI makes synthetic images, realistic videos and cloned voices increasingly difficult to distinguish from authentic media, the ability to establish where digital content came from is becoming a major part of online safety.

Google says its SynthID verification features across Gemini, Search and Chrome are already handling more than 1 million verification requests per day.

How SynthID Works

SynthID is not simply looking at a picture and trying to decide whether it “looks like AI.”

Instead, Google embeds an imperceptible digital watermark into AI-generated content.

The watermark is designed to remain detectable even after common modifications such as cropping, filtering, compression and other forms of editing. For audio, Google says the watermark can also survive changes such as MP3 compression, added noise and changes in playback speed.

When someone uploads supported media to the SynthID Detector, the system searches for that embedded signal.

If it finds one, the detector can indicate that the content — or particular portions of it — are likely to contain a SynthID watermark.

That makes the technology fundamentally different from an AI detector that attempts to judge content purely from its appearance or writing style.

Images, Videos and Audio Are Supported

The new detector covers three major types of media:

Images: JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF and GIF.

Video: MP4, MOV and WEBM.

Audio: WAV, MP3, OGG, FLAC, AAC and M4A.

That makes the tool potentially useful for journalists investigating suspicious footage, researchers studying synthetic media and ordinary internet users who encounter questionable content online.

Google has also integrated SynthID verification into other products, including Gemini and Chrome.

It Is Not a Universal AI Detector

This is perhaps the most important detail for users to understand.

A negative SynthID result does not automatically prove that content was created by a human.

SynthID works by detecting the watermark.

If a piece of content was generated by an AI system that does not use SynthID, the detector may have nothing to find.

Google's own documentation describes SynthID as a way to identify content carrying its watermark rather than a universal system capable of identifying every AI-generated file on the internet.

That limitation becomes particularly important as the AI industry develops multiple competing approaches to content provenance.

More AI Companies Are Joining the System

Google is also trying to make SynthID bigger than a Google-only system.

The company says NVIDIA, OpenAI and Kakao have adopted SynthID, with Apple expected to add support. Google has also previously announced partnerships around expanding SynthID across the AI ecosystem.

That matters because a watermarking system becomes much more useful when content created by different AI companies can be identified through a shared verification mechanism.

Google has also been working with C2PA Content Credentials, another approach designed to provide information about the origin and editing history of digital content.

The broader direction is therefore moving beyond simply asking:

“Is this AI?”

The industry is increasingly trying to answer:

“Where did this content come from, what tools created it and how has it been changed?”

SynthID Has Been Around Since 2023

Google introduced SynthID in 2023 as a watermarking system for AI-generated images created using Imagen.

The technology has since expanded to other forms of media, including video, audio and text.

Google says SynthID has now been used to watermark more than 180 billion images and videos, as well as approximately 240,000 years of audio content.

The company has also continued integrating verification into products people already use.

Gemini can check uploaded media for SynthID, while Google has been expanding content-verification capabilities across Search and Chrome.

Why This Matters for Deepfakes

The problem SynthID is trying to address is bigger than AI-generated artwork.

Synthetic media is increasingly being used to imitate real people, manufacture events and create convincing audio or video that can spread rapidly before anyone verifies it.

A fake video of a public figure can be shared thousands of times before a correction appears.

A fabricated voice recording can be used to impersonate someone.

A realistic image can be presented as evidence of an event that never happened.

Traditional clues are becoming less reliable as generative models improve.

That is why provenance technologies such as SynthID and C2PA are becoming increasingly important.

They do not necessarily tell users whether a piece of content is true.

They can, however, provide additional information about how the content was created or altered.

The Verification Problem Is Getting Bigger

The challenge is that no single technology currently provides a complete solution.

Watermarks can only be detected when they exist.

Metadata can be removed.

Content can be edited.

Different AI companies may use different standards.

And even authentic photographs and videos can be misleading when they are taken out of context.

Google itself acknowledged from the beginning that SynthID was not designed to be foolproof against extreme manipulation.

That means users should treat verification tools as another layer of evidence rather than an automatic truth machine.

A New Tool for Newsrooms and Everyday Users

For journalists, the public launch could be particularly useful.

Newsrooms increasingly receive photographs, videos and audio clips through social media and messaging platforms, often without reliable information about their origin.

A quick watermark check could help establish whether a file was generated using a participating AI system before it is published or used as evidence.

For ordinary users, the same principle applies.

If a suspicious video starts circulating on WhatsApp, X, Instagram or Facebook, users can now test the file rather than relying entirely on how realistic it looks.

That does not guarantee the media is authentic.

But it gives people another piece of information before they share it.

The biggest development here isn't simply that Google has launched another AI detector.

It is that content provenance is becoming part of the infrastructure of the internet.

As AI-generated media becomes easier to create, platforms will need ways to show users where digital content came from and how it was produced.

SynthID is one approach.

C2PA is another.

The real challenge will be getting enough AI companies, platforms and hardware makers to adopt interoperable standards so that verification works across the internet rather than only inside individual ecosystems.

For now, users should remember one rule:

A SynthID result can provide useful evidence about whether supported AI-generated content carries Google's watermark — but no watermark check can, by itself, prove that everything you see online is true.