Google opens SynthID verification site so anyone can check media for AI watermarks

What happened

Google has launched a public website that lets anyone check whether a piece of media — an image, a video or an audio clip — was generated by AI. According to TechCrunch, the site went live on Tuesday, October 7, and turns a capability previously restricted to a small group into something anyone on the web can use.

The tool was not built from scratch for this launch. TechCrunch reports that Google had already given select journalists, media professionals and researchers access for testing around last year's Google I/O, and the new site simply extends that access to the general public.

How the verification works

The site is powered by SynthID, the watermarking technology Google first introduced in 2023. Rather than trying to detect the visual or acoustic fingerprints of synthetic content after the fact, SynthID works by embedding a signal into media at the moment it is generated. Verification then checks whether that signal is present.

Google's own generation stack applies the watermark across the board. TechCrunch notes that the Nano Banana, Veo and Lyria models all use SynthID, as do generation-capable tools such as Gemini, Flow, ProducerAI and Vids.

The site accepts a wide range of formats. For images, it handles JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF and GIF. Video uploads can be MP4, MOV or WEBM, and audio files can be WAV, MP3, OGG, FLAC, AAC or M4A.

Distribution and demand

The website is the most open entry point, but it is not the only one. Google has already built SynthID verification into the Gemini app and into Chrome, meaning many users encounter the capability where they already browse and work rather than having to seek out a dedicated tool.

There is evidence of real demand. According to TechCrunch, Google says people currently make around 1 million requests to verify content every day.

A fragmented landscape with known gaps

SynthID is not a Google-only effort. TechCrunch reports that OpenAI, NVIDIA and Kakao also support the technology, and OpenAI maintains a separate site of its own for checking content. Apple is said to be adding support soon as well.

Not everyone is on board, though. Microsoft and Meta maintain their own standards for watermarking and verifying content, which points to a fragmented provenance ecosystem where a single check may not cover everything.

The technology also has real limitations. TechCrunch highlights that these verification tools are not infallible, and they often fail to identify content created by the very models belonging to the companies that built them. In other words, even a widely adopted watermarking scheme is a detection aid, not a guarantee — a watermark can only be found if the generating model applied it in the first place.

Why it matters

Synthetic media is becoming cheaper and easier to produce, and the ability to check provenance is shifting from a specialist research task to an everyday utility. Google opening SynthID verification to everyone, with support already baked into Chrome and the Gemini app, pushes watermark-based detection toward mainstream reach at a reported pace of a million checks per day.

The cross-industry support from OpenAI, NVIDIA and Kakao, plus Apple's reported plans, suggests SynthID could become a de facto shared standard rather than one company's proprietary scheme — which matters, because provenance only works if it travels across model makers.

The caveats are just as important. Watermarking catches only content generated by participating systems, competing standards leave gaps, and even in-house tools regularly miss their own models' output. The new site is a practical, genuinely useful step against synthetic content, but it is best understood as one layer of defence rather than a complete answer.