TL;DR: Google opened its SynthID Detector to everyone globally in English on Oct. 7, 2026, letting anyone with a supported login check images, video and audio at synthid.com for invisible AI watermarks from Google, OpenAI, NVIDIA and Kakao, with Apple listed as coming soon.

Google opens watermark checker to the public

Google made its SynthID Detector available worldwide in English starting Oct. 7, 2026, according to Google's official blog. The announcement came from Pushmeet Kohli, VP of Science and Strategic Initiatives at Google DeepMind.

The tool lets any user check whether a photo, video or audio file carries a SynthID marker showing it was made or edited with a supported AI system, according to Google's official blog. The company first introduced SynthID in 2023 as a method for placing imperceptible marks in synthetic media, according to TechRepublic.

An early version of the detector was offered last year to media professionals, according to Google's official blog. Journalists, researchers and other media professionals could join a waitlist for that earlier access, according to Search Engine Journal. The Oct. 7 expansion removed that gate and opened the same portal to all users, according to Android Authority.

The public portal runs at synthid.com, where visitors upload a file for inspection, according to Search Engine Journal. Users reach it directly without asking a chatbot to perform the check, according to Ars Technica. The Verge reported that anyone can use it by logging into synthid.com.

What the new site checks

The detector looks for SynthID, described as a near-invisible marker placed in AI-generated material, according to The Verge. It does not judge whether a file looks synthetic, according to TechRepublic. It instead searches for the hidden signal inserted by a participating generation system, according to TechRepublic.

For pictures and clips, that signal is encoded in pixels, according to Ars Technica. For sound, it is encoded in the waveform, according to Ars Technica. The marks are designed to remain detectable after common changes, according to TechRepublic. As an example, audio watermarks can survive noise, MP3 compression and changes in playback speed, according to TechRepublic.

The expanded detector supports distinct watermarks from Google and its SynthID partners, according to Ars Technica. Named partners are OpenAI, NVIDIA and Kakao, according to Android Authority. Apple support is described as coming soon, according to Android Authority. MacRumors reported that Apple plans to support SynthID later in the year.

That cross-company coverage is a change from earlier Google checking options, according to Android Authority. Gemini could only identify marks created with Google AI tools, while OpenAI's verification site could only identify signatures created with OpenAI tools, according to Android Authority. Ars Technica reported that ChatGPT images carry SynthID but were not flagged in Google's earlier detector.

The result tells a user whether SynthID was used to produce or edit the media, according to Ars Technica. Google's post presented a sample video result stating "SynthID was detected" and "This media was made or edited with Google AI," according to Search Engine Journal. That result also noted the media "may have been edited further since then," according to Search Engine Journal.

The public site does not separate creation from editing, according to MacRumors. Google's tool cannot distinguish between AI-created and AI-edited media, according to MacRumors. Search Engine Journal noted that a photo touched up with Google AI and one generated from scratch with it might both receive the same "made or edited" label.

The internal tool used by trusted testers offered more detail by highlighting regions that showed SynthID pixels, according to Ars Technica. The public website does not provide that regional view, according to Ars Technica. Public users receive a message stating whether SynthID was used to produce or edit the file, according to Ars Technica.

How to use synthid.com

Access requires sign-in, according to MacRumors. People can log onto synthid.com with a Google, Apple or OpenAI account, according to Android Authority. Ars Technica reported that the tool supports Google, OpenAI and Apple accounts.

TechRepublic described the workflow as uploading media and checking whether it contains an invisible AI watermark. Search Engine Journal reported that the site asks users for the best-quality copy of a file they have, saying that gives the most accurate result.

Each user receives what Google describes as "a daily usage quota of approximately 10 image, video, and audio checks," according to Ars Technica. Google engineers told Ars Technica that limiting access is necessary to prevent people from using the checks to develop SynthID removal tools. Ars Technica reported that uncertainty around the exact number likely reflects that the system may restrict users sooner when they check multiple very similar images, a pattern that could indicate tuning of a bypass workflow.

The website handles common consumer formats, according to TechRepublic. It supports 11 image formats, three video formats and six audio types, according to TechRepublic. Listed examples include JPG, PNG, WEBP, GIF, MP4, MOV, MP3, WAV and FLAC, according to TechRepublic.

Google also offers verification outside the dedicated portal, according to MacRumors. SynthID checks are already integrated into Search, Gemini and Chrome, according to MacRumors. Google said those built-in features now regularly handle over 1 million requests daily, according to Google's official blog.

Scale of watermarking so far

Google said it has watermarked over 180 billion images and videos to date, according to Google's official blog. Ars Technica reported that Gemini alone has integrated SynthID with over 180 billion images and videos.

For audio, the total is described as 240,000 years of content, according to Google's official blog. TechRepublic reported the same audio total alongside the image and video total.

The company said SynthID watermarks are added across generative AI consumer products from participating companies, according to Search Engine Journal. The detector site states: "This is not a general AI detector. We can only detect media from the above companies, who have adopted SynthID technology," according to Search Engine Journal.

What a negative result does not prove

The site's FAQ explains that "no signal is completely robust" and advises "collecting multiple points of data before making a decision," according to Search Engine Journal. That caution matters because many AI outputs fall outside the system, according to Ars Technica.

Some model providers use different watermarking technology, according to Ars Technica. Meta was cited as an example of a provider using different technology, according to Ars Technica. Other providers use no watermarks at all, according to Ars Technica.

Open-weight models present another gap, according to Ars Technica. People can run those models on their own hardware to produce increasingly realistic material without simple metadata labeling, according to Ars Technica. That material will not be identified as AI in the SynthID tool, according to Ars Technica.

Even when a file comes from a participating company, removal remains a concern, according to Ars Technica. Ars Technica reported it is plausible that someone will find a way to remove SynthID, making existing AI content look more legitimate. TechRepublic described the central limitation in similar terms: no watermark does not mean no AI.

Ars Technica noted the website will carry a disclaimer to that effect. The publication also discussed cryptographic alternatives such as C2PA for labeling real captures at creation, according to Ars Technica. Google's Pixel phones are among the few devices doing that, according to Ars Technica.

Why Google expanded access now

Google framed the release as a response to more capable and easier-to-use AI tools, according to Google's official blog. The company said clear context about online content helps people make informed decisions about what to trust, according to Google's official blog.

The tool joins verification features already present in Search, the Gemini app and Chrome, according to Google's official blog. Google described the broader effort as giving users more context about media seen online so they can navigate the web with confidence, according to Google's official blog.

Android Authority called the public portal a step forward for media verification in the age of generative AI. The same report noted the timing followed Anthropic's announcement that Claude can add watermarks to AI-generated text, according to Android Authority.

For newsrooms, researchers and everyday users, the practical effect is a single destination for checking files linked to popular AI tools, according to Ars Technica. Before the change, outsiders outside Google's trusted tester program had to ask Gemini, a process Ars Technica described as clunky and sometimes confusing when users wanted a simple yes-or-no answer.

Sources referenced