Google Opens SynthID AI Detector to Everyone, Letting Users Check Images, Videos and Audio
 |
| Image: Google |
Google has made its SynthID Detector available to the public, giving users a new way to check whether images, videos and audio files contain invisible watermarks associated with AI-generated content. The tool, which was previously tested with media professionals, is now available globally in English and supports compatible content from Google and several partner companies.
The launch comes as AI-generated media becomes increasingly realistic and harder to distinguish from authentic photographs, recordings and videos. Rather than judging content by how it looks or sounds, SynthID Detector searches for digital watermarks embedded in supported AI-generated media.
However, the tool has an important limitation: it is not a universal AI detector, and failing to find a watermark does not prove that a file was created by a human.
What Is Google SynthID Detector?
SynthID Detector is an online verification tool developed around Google DeepMind's SynthID watermarking technology. It allows users to upload supported media and check whether the content contains a recognizable SynthID watermark.
Google introduced SynthID in 2023 to help identify AI-generated content through watermarks designed to remain imperceptible to people. The technology can embed these signals into images, videos and audio without adding a visible logo or an audible marker.
The public detector expands access to this verification process beyond the media professionals who received access to an earlier version in 2025.
Users can access the service through the official SynthID Detector website.
How Does SynthID Detector Work?
Traditional watermarks are often visible logos or text placed over an image. SynthID takes a different approach by embedding information into the underlying media signal.
When a user submits a compatible file, the detector searches for a watermark that its system recognizes. If it finds one, the result can help establish a connection between the content and an AI system that uses the technology.
The process works across three main types of media:
- Images: The system checks visual content for a compatible watermark.
- Videos: It can examine video content for watermark signals, helping users investigate potentially AI-generated clips.
- Audio: It checks supported recordings for watermarks embedded in the audio signal.
The technology is designed to withstand certain common modifications, although detection is not guaranteed in every situation. Heavy editing or other changes may affect whether a watermark can be identified.
Importantly, SynthID does not independently determine the complete history of a file. Finding a watermark provides evidence of a connection to a supported AI-generation system, but it does not automatically establish whether the entire file is synthetic, whether it has been edited, or whether the claims made alongside it are true.
Which AI Platforms Does SynthID Detector Support?
Google says the public detector can check compatible media generated using its own AI technology and systems from partner companies.
At launch, the announced partners included:
- OpenAI
- NVIDIA
- Kakao
Apple was listed as a future addition in Google's announcement on October 7, 2026.
This expansion is significant because it brings several AI companies into the same watermark-verification ecosystem. Nevertheless, support for a company does not mean that every output from its products will carry a detectable SynthID watermark.
Detection depends on the particular system, the content produced and whether a compatible watermark is present and recognizable.
Google Says SynthID Has Watermarked Billions of AI Images and Videos
Google reports that SynthID has been used to watermark more than 180 billion images and videos, along with audio content equivalent to 240,000 years, since the technology launched in 2023.
The company also says verification features integrated into Google Search, the Gemini app and Chrome regularly handle more than one million requests per day.
These figures illustrate the scale of Google's watermarking efforts, although they are company-reported statistics rather than an independent count of all AI-generated media available online.
The public release of SynthID Detector adds a dedicated destination for users who want to check a file directly instead of relying exclusively on verification features integrated into other services.
How to Check Whether an Image, Video or Audio File Was Made With AI
Using the detector involves a straightforward process:
- Visit the SynthID Detector website.
- Open the upload interface and select the supported image, video or audio file you want to examine.
- Submit the file for analysis.
- Review the result to see whether the detector identifies a compatible watermark.
The tool is intended to help users investigate suspicious media, including viral social-media images, manipulated-looking videos and audio recordings whose origins are uncertain.
Availability and supported file formats may vary, so users should consult the service itself for current upload requirements.
A positive result can provide useful evidence about a file's origin. It should not be treated as proof that every element of the media is AI-generated or that its accompanying description is accurate.
Can SynthID Detect Every AI-Generated Image or Video?
No. This is the most important limitation to understand.
SynthID Detector looks for compatible watermarks rather than attempting to recognize every possible form of AI-generated content. If a file was produced using a system that does not embed a supported watermark, the detector may not identify it as AI-generated.
A negative result can therefore have several explanations: the content may be human-created, the generating system may not use a supported watermark, or the watermark may not be detectable in the submitted version.
The distinction matters when checking potentially misleading content. A realistic AI-generated image could pass a watermark check simply because it does not contain a watermark the service recognizes.
Likewise, finding a watermark does not establish that an image is misleading, harmful or factually incorrect. AI-generated content can be legitimate, while authentic photographs can also be presented in misleading contexts.
For these reasons, watermark verification works best alongside other checks, including tracing the original source, comparing independent reports and examining the context in which the content appeared.
Why Google's SynthID Detector Matters
Making watermark verification publicly accessible could help journalists, researchers, content creators and ordinary internet users investigate the origins of digital media.
The approach also highlights a broader challenge facing the technology industry: identifying AI-generated content is not necessarily the same as determining whether content is trustworthy.
Watermarking offers one way to establish provenance, but its effectiveness depends on adoption across AI platforms and the ability to detect embedded signals after media has been shared or modified.
Google's expanded detector represents a step toward more accessible verification, while the participation of other AI companies could help extend the reach of compatible watermarking systems.
Still, no single verification tool can establish the authenticity of every image, video or recording circulating online.
The Bottom Line
Google's public release of SynthID Detector gives users a convenient way to check images, videos and audio for watermarks associated with supported AI systems. Its expansion to include compatible content from Google, OpenAI, NVIDIA and Kakao makes it relevant beyond Google's own AI products.
However, the detector should be understood as a watermark-verification service rather than a definitive test of whether something was made by AI. A detected watermark can provide valuable evidence, but an absent watermark does not establish that content is genuine.
As AI-generated media continues to spread, combining watermark checks with source verification and independent evidence remains essential.
