We’re Making It Easier To Identify AI-generated Content Globally.

TL;DR

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Google has opened its SynthID Detector to the public worldwide in English. The tool can check images, video and audio for SynthID watermarks associated with Google and participating partners, but a check does not establish whether all AI-generated content can be identified.

Google has opened its SynthID Detector to anyone worldwide using it in English, expanding a tool that checks images, video and audio for watermarks linked to AI-generated media. Google says the detector can identify content made with AI from Google and participating partners, including OpenAI, NVIDIA and Kakao, with Apple support planned soon.

The detector is an expansion of an early version introduced last year for media professionals. Google says users can submit an image, video or audio file to check for SynthID. The announcement describes the tool as a way to identify watermarks, not as a general test that can determine whether any media was generated by AI.

Google says it has applied SynthID watermarks to more than 180 billion images and videos and to audio representing 240,000 years. Those figures are the company’s cumulative totals; the announcement does not give a measurement period, independent audit or comparison baseline for them.

The detector adds to verification features Google says are built into Search, the Gemini app and Chrome. The company reports that those features handle more than one million requests a day. That usage figure refers to Google’s existing features, not necessarily requests made through the newly expanded detector.

A Wider Check for Watermarked Media

Public access gives people beyond media organisations a way to check for specific AI-content markers in media they encounter. That may help users seeking context about an image, clip or recording, and could make watermark checks easier to incorporate into reporting and other verification work.

The tool’s usefulness depends on whether the content carries a watermark the detector recognizes. Google’s announcement covers SynthID-marked material from its own systems and named partners; it does not say the detector can identify every AI-generated file, or media made with systems that do not apply compatible watermarks. A result should therefore be understood as information about a particular watermark, rather than a complete verdict on an item’s origin.

From Professional Preview to Public Access

Google introduced SynthID in 2023, describing it as an imperceptible watermarking system for AI-generated images, video and audio. The company says watermarks are embedded in media and can be checked using its detection technology. It introduced an early detector version for media professionals last year; the latest announcement expands access beyond that initial group.

Google also describes the detector as part of a wider set of efforts to provide context about online media, including verification features in Search, Gemini and Chrome. The announcement does not detail how those features determine or display results, so their operation should not be assumed to be identical to the new standalone detector.

““With the new SynthID Detector, anyone can easily check if an image, video, or audio file was made with AI from Google or our partners.””

— Google, in its announcement

Limits of the Detector’s Coverage

Google has not said that SynthID Detector can identify all AI-generated media. The announcement limits its stated coverage to images, video and audio that carry watermarks it can recognize, and names Google, OpenAI, NVIDIA and Kakao as participating sources, with Apple support expected soon. It does not explain whether the detector can identify material from other tools or media whose watermark has been altered or removed.

The company has also not provided details in this announcement on the detector’s accuracy, how it handles partial or inconclusive matches, file-size limits, privacy and data retention, or whether results include an explanation of the detected watermark. The release date is described as “today” without a calendar date, and the announcement does not give a timeline for Apple support or for availability in languages other than English.

Partner Support and Language Rollout

Users can access the detector globally in English, according to Google. The company says Apple support is coming soon, but has not supplied a date. It has not announced when the service may support additional languages or whether further media providers will join.

As access expands, the practical questions will include which watermarks the tool can recognize and what information it returns when a match is not clear. Until Google publishes more technical and operational detail, the detector is best treated as one check for supported watermarks—not a definitive way to establish the origin of every image, video or audio recording.

Key Questions

What is Google’s SynthID Detector?

It is a tool that Google says checks images, video and audio for SynthID watermarks associated with AI-generated content from Google and participating partners.

Who can use the detector?

Google says the detector is now available to anyone globally in English. The company does not provide a calendar date for the announcement in the source material.

Which companies’ AI content can it check?

Google names Google, OpenAI, NVIDIA and Kakao as participating sources. It says Apple support is planned soon, without giving a launch date.

Can it identify every AI-generated image or recording?

Google does not claim universal coverage. The announcement describes checks for watermarks the detector recognizes, so it should not be treated as a definitive test for media made by every AI system.

What is still unknown about the service?

The announcement does not specify the detector’s accuracy, data-handling practices, results for uncertain matches, support beyond English, or the schedule for Apple and other future partners.

Source: rss