The tool, built around Google's existing SynthID watermarking technology, went live Tuesday and is open to the general public rather than limited to enterprise partners or researchers. Users can upload a piece of media and the site will scan it for an embedded digital signature that reveals whether it was produced by AI generation tools, even after the file has been edited, compressed, or otherwise altered.
SynthID itself isn't brand new. Google has spent the past couple of years embedding the invisible watermarking system into content produced by its own generative AI tools, including image and video models. What's changed is accessibility: instead of the detection capability living behind developer APIs or internal dashboards, it's now packaged as a consumer-facing verification site that anyone with a suspicious file can use in seconds.
The timing matters. AI-generated images, deepfake videos, and synthetic voice clips have become sharply more convincing and far cheaper to produce over the last two years, fueling concerns about misinformation, political manipulation, scams, and fraud. Platforms and regulators alike have struggled to keep pace, and watermarking has emerged as one of the few technical approaches with broad industry backing as a way to label synthetic content without relying purely on after-the-fact fact-checking.
The approach does have real limits. SynthID can only flag content that was watermarked at the point of creation, meaning it offers no help with AI media generated by tools that don't use the system, or with watermarks that have been deliberately stripped or degraded through heavy editing. That leaves a significant share of the deepfakes circulating online outside its detection net, since watermarking adoption is still far from universal across the AI industry.
Even so, giving ordinary users direct, free access to a provenance-checking tool is a notable shift from a landscape where media authenticity verification has largely been the province of specialized forensic software or newsroom tools. If other major AI developers adopt compatible watermarking standards, tools like this could eventually form the backbone of a broader, industry-wide system for labeling synthetic content at scale. For now, it gives everyday internet users a new, if imperfect, first line of defense against being fooled by AI-made fakes.