Key Notes
- Google opens SynthID Detector globally in English after an earlier media-focused trial.
- The portal checks supported watermarks in images, video and audio.
- A missing watermark does not prove that a file is authentic.
Google has opened SynthID Detector to everyone, giving the public a dedicated website for checking AI watermarks in images, video and audio. The service is available globally in English, expanding beyond the early-access version introduced for media professionals in 2025.
The October 7 announcement names Google, OpenAI, NVIDIA and Kakao as supported providers, with Apple support coming later. The wider access makes a useful verification tool easier to reach, but its results need careful interpretation: it looks for supported watermarks, rather than establishing whether every possible file was made by AI.
What SynthID Detector Checks
SynthID works by embedding an imperceptible signal when participating AI systems create content. Google DeepMind’s technical overview describes watermarks in image pixels and video segments, with a separate approach for audio. The detector looks for those signals after someone submits media for verification.
That is a different question from asking whether an image simply looks artificial. A convincing picture may carry a detectable watermark, while an obviously synthetic picture from an unsupported generator may not. A missing signal therefore leaves the file’s origin unresolved; it does not certify that a camera recorded the scene.
Google says its image and video watermarks are designed to survive common changes such as cropping, filters and compression. That is a robustness goal, rather than a reason to assume every edited or repeatedly re-encoded copy will return an equally clear result. When possible, checking a file closer to its original version should preserve more of the evidence available for inspection.
Images, Video and Audio in One Place
The public portal gives users a direct place to check media, alongside verification features Google already provides in Gemini, Search and Chrome. According to the company, those built-in features regularly handle more than 1 million requests a day.
The scale of watermarking is also growing. Google says it has marked more than 180 billion images and videos, plus audio equivalent to 240,000 years. Those are the company’s cumulative figures, rather than a count of files that users have independently verified through the new public website.
For people experimenting with tools such as Nano Banana, the rollout creates a more accessible way to investigate media provenance. It does not make a separate assessment of whether a caption is accurate, whether a depicted event happened or whether someone has attached a genuine image to a misleading claim.
How to Interpret a Detection Result
A positive watermark result can contribute evidence about the creation process. It still needs to be read alongside the source of the file, its publication history and the claim being made about it. For example, confirming that an illustrative image involved AI would not by itself establish whether the accompanying news report is true or false.
The same caution runs in the other direction. A photograph without a detected watermark could be authentic, could come from an unsupported AI system, or could have passed through changes that affect detection. Treating all three possibilities as “real” would turn a limited technical finding into a much broader conclusion than it supports.
Watermarking also varies by medium. Our reporting on text provenance concerns generated writing; this public rollout is focused on images, video and audio. Readers should check which type of content a verification service actually supports before relying on its output.
What Happens to Uploaded Files?
The portal’s privacy notice says uploaded media is processed in real time and deleted once detection results are returned. It separately states that a digital signature of the media is linked to the user’s identifier for quota enforcement and abuse prevention, then deleted after 24 hours.
The service’s additional terms prohibit using the frontend as an automated API for bulk verification. For individual checks, the practical value is straightforward: people can now look for supported AI watermarks without belonging to the earlier media-testing program. The result is one more piece of evidence to assess, alongside the source and context of the content.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
