On October 7, 2026, Google DeepMind took its AI watermark checker public. The SynthID Detector — previously available only to journalists, researchers, and media professionals — is now open to everyone at synthid.com, globally, in English. Upload any image, video, or audio file, and the tool tells you whether it carries an invisible watermark from a major AI system.
For creators, freelancers, and small businesses, this matters more than it looks. The internet is flooding with synthetic media, and "does this look real?" is no longer a strategy. Here's what Google actually launched — and how to use it.
What the SynthID Detector does
SynthID is Google DeepMind's watermarking technology, first introduced in 2023. It embeds an imperceptible digital signal directly into the pixels of images and video — and into the waveform for audio — so AI-generated content can be identified after the fact. The signal is designed to survive common transformations like cropping, compression, and resizing.
The public Detector checks your upload against watermarks from Google, OpenAI, NVIDIA, and Kakao, with Apple expected to join soon. One upload checks multiple vendors at once — a meaningful upgrade from the previous fragmented setup, where you had to route files through Gemini or visit each company's own checker.
The scale Google cites is enormous: since 2023, SynthID has watermarked more than 180 billion images and videos and the equivalent of 240,000 years of audio. The public Detector joins verification features already built into Google Search, the Gemini app, and Chrome, which Google says now handle more than one million checks per day.
The hard limits: what it does NOT tell you
Pushmeet Kohli, Google DeepMind's VP of Science and Strategic Initiatives, summed up the launch in the official announcement: "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, including OpenAI, NVIDIA, Kakao, and soon, Apple."
But read the fine print before you treat it as a lie detector:
- It is not a general AI detector. It hunts for a specific digital signature. A file with no watermark gets no alert — and that tells you nothing about whether a human made it. Media from tools that don't use SynthID slips through.
- It can't tell "fully AI-generated" from "lightly edited." A detected watermark means the file was "likely edited or generated" by a SynthID-enabled model — a real photo touched up with a Gemini edit can read the same as a fully synthetic image.
- Access has friction by design. You must sign in with a Google, OpenAI, or Apple account, there's a daily usage quota (around 10 checks per user per day, per Ars Technica), and the terms ban scripts and automated bulk submissions. It's built for one-file-at-a-time manual checks.
- Uploads are transient. Uploaded files are deleted once results return — useful to know before checking client material.
Supported formats
Google's portal accepts common formats: JPG, PNG, WEBP, HEIC, and GIF for images; MP4, MOV, and WEBM for video; WAV, MP3, FLAC, and others for audio. Upload the best-quality copy you can find — reposts and screenshots can weaken the watermark.
What this means for you: a creator's playbook
Here's where this tool pays for itself.
1. Verify assets before you pay for them
Buying stock-style images, hiring a thumbnail designer, licensing music? Upload the file first. A watermark match means AI was involved somewhere in its creation — that's a data point for negotiating price and for setting client expectations. Nobody wants to discover their "custom brand shoot" was a Gemini edit after the invoice is paid.
2. Check suspicious viral content before you share it
Creators share and react to viral media constantly — and sharing a fake that damages someone's reputation can cost you an audience, a client, or worse. A 10-second SynthID check before you quote-post a scandalous clip is the cheapest insurance in your workflow.
3. Protect your own work with provenance
The flip side: if you generate or edit with SynthID-enabled tools, your content now carries a verifiable mark. That's an advantage in a market where clients increasingly ask "is this real?" — you can prove your pipeline, and your licensing terms can reflect the tooling behind the work.
4. Don't outsource judgment to it
The biggest risk is over-trusting the tool. A clean result is not a certificate of authenticity — it only means no SynthID watermark was found. Keep it as one signal among others: source reputation, metadata, reverse image search, and your own judgment.
The bottom line
Google didn't build a universal AI detector — no such thing exists yet, and the company is explicit about that. What it built is a cross-vendor watermark checker that finally works for ordinary users, free, in a browser. For a creator economy running on trust, a free 10-second provenance check is a tool worth having in the belt. Use it, understand its limits, and never let a clean result be the last word.
Watch Google DeepMind's official explainer of how SynthID watermarking works:
Sources
- SynthID Detector — official portal: the public watermark checker, launched October 7, 2026
- Google DeepMind — SynthID: official Google DeepMind video explaining SynthID watermarking