Google’s new SynthID website can identify AI-generated media

What happened

Google has launched a public website built on its SynthID technology, allowing anyone to check whether an image, video or audio clip was generated or edited using AI. Users can upload a file to the site, which scans for an invisible SynthID watermark embedded by Google's generative AI tools and reports back on whether AI involvement is detected.

SynthID was originally developed as a behind-the-scenes watermarking system, embedding imperceptible signals into content produced by Google's own AI models. The new standalone site makes that verification capability directly accessible to the public for the first time, rather than keeping it confined to enterprise or developer tools.

Why it matters

The launch reflects a broader shift in how AI developers are being pushed, and are choosing, to address the provenance problem: as generative tools make convincing synthetic images, video and audio trivially easy to produce, the ability to verify origin becomes a practical necessity rather than a nice-to-have. By putting a verification tool directly in the hands of the public, Google is signalling that detection infrastructure needs to be as accessible as generation infrastructure.

For organisations building digital trust, content moderation or compliance workflows, this points to watermarking and verification becoming a standard layer of the AI stack, not an afterthought. It also raises the practical limitation that SynthID can only detect content made with tools that actually embed its watermark, meaning coverage is partial rather than universal.

The Renascence take

Tools like SynthID matter less as a technical fix and more as a signal of where public expectations are heading. The real test isn't whether the watermark works in a lab demo, it's whether ordinary users will actually reach for a verification step before they trust what they see.

Verification tools only change behaviour if checking becomes as frictionless as sharing. Right now, trust still defaults to the loudest or most emotionally compelling content, not the most verified. Service and platform designers should treat provenance checks less as a standalone utility and more as something embedded at the point of consumption, inside feeds, inboxes and messaging apps, where the moment of doubt actually happens. Anything that asks people to detour to a separate website will be used by the curious minority, not the vulnerable majority who need it most.