Google just opened its SynthID Detector to the public, giving anyone a tool to check whether an image, video, or audio file was AI-generated—and handing the same company caught rigging search results and shadow-banning conservatives another lever over what Americans are allowed to trust online.
The stakes are straightforward. AI-generated content is flooding the internet, and ordinary people need ways to tell what's real. But the company that spent years throttling dissenting voices and burying unwanted search results now positions itself as the honesty police. If Google controls the detection infrastructure, Google controls the definition of "authentic."
Here's what SynthID Detector actually does, according to Engadget: it scans uploaded files for an imperceptible watermark that Google's Gemini AI model has been embedding in everything it creates since 2023. You go to SynthID.com, upload a file, and get a yes-or-no answer on whether a watermark is present. The tool also checks watermarks from partner companies including OpenAI, NVIDIA, and Kakao. Apple is a SynthID participant but detector support for its content isn't ready yet.
The simplicity is by design—and it's also the trap. Because SynthID Detector only checks for watermarks, it cannot detect AI content from any model that doesn't participate in the SynthID system. Meta's AI tools, for instance, use their own watermarking system called Content Seal, which Reuters found in 2026 sometimes failed to accurately identify AI-generated material. Any open-source model, any foreign tool, any bad actor who simply skips the watermark—SynthID Detector returns a clean bill of health. A "no" doesn't mean the content is real. It means Google didn't stamp it.
Google is also keeping the leash tight. Users are limited to roughly 10 checks per day, a restriction a Google spokesperson told Ars Technica is meant to stop people from reverse-engineering the watermark and stripping it out. You also have to be logged into a Google, OpenAI, or Apple account to use the tool at all—so every query feeds data back to one of the tech oligopoly's platforms.
Engadget framed the rollout as a positive step—"one of the most reliable" detectors available—while burying the core problem: a detection tool that only catches its own cartel's output isn't a transparency measure. It's a walled garden. The Sioux City Journal, for its part, had nothing on the story at all, running instead a local piece on a Google memorandum of understanding with the town of Salix—corporate outreach to local government while the real power play happens at the infrastructure level, unnoticed.
The open question is who watches the watchmen. Google builds the AI, Google watermarks the AI, and Google builds the detector that certifies what's real. That's not a check on power. That's a vertical monopoly on truth.
