Google's SynthID Detector Is Public: What Sellers and Creators Should Do

On Oct. 7, 2026, Google opened its SynthID Detector to everyone, globally and in English, so anyone can check whether an image, video or audio file carries a SynthID watermark from Google or its partners OpenAI, NVIDIA and Kakao, with Apple support coming, according to Google. Google says it has already watermarked more than 180 billion images and videos and 240,000 years of audio.

That matters well beyond newsrooms. If you sell online or make product content with AI tools, your shoppers, competitors and the platforms you sell on now have a free, simple way to check your images. Here is what launched, what it can and can't tell anyone, and what sellers and creators should do this week.

What did Google launch?

The detector lives on a dedicated site, synthid.com, The Verge and Engadget reported. It is free, but you have to sign in with a Google, OpenAI or Apple account, and Google says each user gets "a daily usage quota of approximately 10 image, video, and audio checks," per Ars Technica. Google engineers told Ars the cap is there to stop people from using repeated checks to build watermark-removal tools.

TechCrunch reported that the site accepts common image formats (JPG, PNG, WEBP, AVIF, HEIC, GIF and more), MP4, MOV and WEBM video, and most audio formats. Google's own Nano Banana, Veo and Lyria models, plus tools such as Gemini, Flow and Vids, watermark everything they generate. Google says its built-in checks in Search, the Gemini app and Chrome already handle more than 1 million verification requests a day.

The answer is deliberately simple. Ars reported that the public site tells you whether SynthID was used to produce or edit the media, without the region-by-region highlighting of Google's internal tool. Until now, people outside Google's tester program mostly had to ask Gemini, and Google's older detector only recognized Google's own watermark.

What can't it tell you?

A "no watermark" result does not mean an image is real. Ars noted that some providers use different watermarking (Meta, for example), others use none, and open-weight models run at home can produce realistic images with no label at all. The site carries a disclaimer saying as much. TechCrunch added that Microsoft and Meta run their own standards, and that such tools often fail to catch content from their makers' own models.

So the detector is a one-way signal. A hit is strong evidence that a supported AI tool made or edited the file. A miss proves nothing.

Why does it matter for sellers and creators?

Three reasons.

First, the platforms already ask for this. Google Merchant Center requires every image created with generative AI to carry the IPTC metadata tag DigitalSourceType TrainedAlgorithmicMedia, and tells sellers not to remove that embedded tag. AI-written titles and descriptions have to go in separate structured_title and structured_description attributes marked as AI-generated.

Second, video platforms label it. YouTube requires creators to disclose realistic AI-generated or meaningfully altered content through the "AI use" setting in YouTube Studio. It may label videos automatically if they contain C2PA metadata or its systems detect AI. YouTube says disclosure won't limit a video's audience or its eligibility to earn money, but creators who keep skipping it risk labels, removals or suspension from the Partner Program.

Third, shoppers can now check for themselves. A lifestyle photo made with a supported tool will show up as AI-made when someone uploads it to synthid.com. That is fine if your listing is honest about it. It is a problem if the image makes the product look different from what arrives.

What happens next?

• Apple support is listed as "soon," per Google, which would widen what the detector recognizes.
• Expect more marketplaces and ad platforms to read AI labels and metadata automatically, as YouTube already does.
• Ars pointed to C2PA content credentials, which label real photos at capture (Google's Pixel phones do this), as the complementary approach to watch.

What should sellers and creators do this week?

1. Make a list. Mark every product image and video as real photo, AI-edited or AI-generated. You can't disclose what you haven't tracked.
2. Spot-check a few files on synthid.com. With about 10 checks a day, test one file from each tool you use so you know which ones watermark.
3. Keep the metadata. Export settings that strip IPTC or C2PA data can put you out of line with Merchant Center rules. Check your image compressor and CMS too.
4. Keep the product true. AI backgrounds, lighting and scenes are one thing; changing the product's color, size, texture or what's in the box is how returns and complaints start. Keep a real photo of every SKU on file.
5. Disclose where it's required. Turn on YouTube's "AI use" setting for realistic AI scenes, and follow each marketplace's own image rules. We covered Amazon-side image and compliance checks in our Listimo seller toolbox guide.
6. Don't use it to accuse. A watermark on a competitor's image is not proof of wrongdoing, and no watermark is not proof of a real photo.

Disclosure: some links are affiliate links; ShopperCove may earn a commission at no extra cost to you.

If you use AI to build Amazon listing images, pick a tool that is upfront about it. Listimo's sales page says its images and videos are AI-generated, tells sellers to review every one before uploading and to disclose AI content wherever a marketplace requires it, and says it is built to keep the product unchanged while generating new surroundings, light and scenes. It doesn't say whether its outputs carry SynthID, so run a test file through the detector yourself. ShopperCove did not test Listimo for this article. Listimo link: Open linkShow address (affiliate link)

Tools that help

• Listimo, Amazon listing AI. Per its sales page, the Seller Pack is 1,800 credits for $229 as a one-time payment ("around 10 complete listings"). It includes a free ASIN listing check, a main-image workshop that rebuilds the main image on pure white, and a statement that it does not guarantee Amazon will accept generated content. Listimo link: Open linkShow address

How this was made: facts were gathered on 8 Oct 2026 (IST) from Google's announcement, the reports and the Google and YouTube help pages listed below, then written up with AI help and checked line by line against those sources. ShopperCove did not run its own SynthID checks for this article.

Sources

• Open blog.googleShow address
• Open arstechnica.comShow address
• Open techcrunch.comShow address
• Open theverge.comShow address
• Open engadget.comShow address
• Open support.google.comShow address
• Open support.google.comShow address
• Open listimo.aiShow address

Related

• Open shoppercove.comShow address
• Open shoppercove.comShow address
• Open shoppercove.comShow address
• Open shoppercove.comShow address