Google SynthID Detector Opens to Everyone to Spot AI Media

Google’s SynthID Detector at synthid.com is open to all. See which files it scans, which AI firms it supports and what a clean result does not prove.

Updated

Share this article

Home › Tech, AI and Business

Google’s SynthID Detector is now open to everyone. This week Google opened public access to the tool at synthid.com, letting anyone upload an image, video or audio clip and check whether it carries the invisible SynthID watermark that Google and several partners embed in AI-generated media, according to TechCrunch and The Verge.

Part of our Tech, AI and Business coverage. Browse more in Artificial Intelligence.

Advertisement

What the SynthID Detector checks

Google first gave the detector to a limited group of journalists, media professionals and researchers after showing it at Google I/O last year. The public version accepts a long list of formats, TechCrunch reported: JPG, PNG, WEBP, AVIF, HEIC, TIFF and GIF images; MP4, MOV and WEBM video; and WAV, MP3, OGG, FLAC, AAC and M4A audio.

The detector looks for SynthID, a watermarking system Google introduced in 2023 that hides a near-invisible marker inside generated content. Google says its Nano Banana, Veo and Lyria models, along with products such as Gemini, Flow, ProducerAI and Vids, stamp their output with it. That includes tools we covered when Google Vids added free 1080p AI video.

Which AI companies support SynthID

SynthID is no longer a Google-only signal. TechCrunch reported that OpenAI, NVIDIA and Kakao also support the watermark, and Apple is said to be adding support soon. OpenAI keeps its own verification site as well. Google has already built SynthID checks into the Gemini app and Chrome, and the company says people make about 1 million verification requests a day.

Engadget, in its coverage of the launch, noted the site is available globally in English. Google described the expansion on its official blog.

The limits of AI watermark detection

A clean result does not prove a file is human-made. The detector can only find the SynthID watermark, so content from tools that use other systems, or no watermark at all, can slip through. Microsoft and Meta run their own watermarking and verification standards, and TechCrunch pointed out that such tools often fail to flag content made by their own makers’ models. Cropping, re-recording and heavy editing can also weaken hidden marks.

Advertisement

That gap matters as synthetic voices and video get more convincing. Boston startups are working on the same problem from the audio side, as in our report on Modulate’s $25 million raise for voice deepfake detection.

Why the SynthID Detector matters for readers

For everyday users, the public site turns a behind-the-scenes standard into something you can actually check before sharing a viral clip. It is most useful as one signal among several: look at the source, look for the original upload and compare with trusted outlets. As more companies sign on, the share of AI media that can be verified this way should grow.

Sources

How to use the SynthID Detector, and what it will not tell you

Engadget reports the tool is free but requires users to log in with a Google, OpenAI or Apple account. Visitors upload a file at synthid.com to check whether it carries a SynthID watermark. Engadget adds that the site will be one of the first detection tools able to identify images made with Apple’s AI features such as Image Playground, with support for Apple Intelligence images coming “soon.”

Engadget also lists the limits. The detector cannot catch all AI-generated content because companies rely on different watermarking standards, watermarks can be stripped in some situations, and the site does not distinguish content that was entirely AI-created from content that was merely edited with AI. It therefore will not say how a file was made or which parts may be synthetic. Google has other ways to check for SynthID watermarks and C2PA content credentials, including Circle to Search and the Gemini app, but a dedicated website is easier to reach for people who do not already use Google’s AI tools.

The scale of SynthID, according to Google

In its announcement, Google DeepMind’s Pushmeet Kohli said SynthID has watermarked more than 180 billion images and videos, along with 240,000 years of audio, since launching in 2023. The post says the new detector is available globally in English and joins built-in verification in Search, the Gemini app and Chrome, which now regularly handle more than 1 million requests a day. Google frames the effort as giving people “more context about the media you see online.”

Further Reading

Read next in Artificial Intelligence

Get The Boston Report Briefing

Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.

Free. We never sell your address. Privacy policy

Get The Boston Report Briefing
Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.
Sign up