Google's New AI Detector Has One Big Blind Spot: Most of the AI You Actually See

Google has officially launched SynthID Detector, a free tool that scans images, video, and audio for an invisible watermark embedded in content created with Gemini and a handful of partner platforms. Sign in with a Google account, drag an image onto the page, and it’ll tell you what it finds, including whether an AI image was edited after it was generated. It’s a genuinely useful idea in a feed full of AI slop.

The catch is who’s missing from the guest list. The detector only recognizes SynthID watermarks, which means images from Midjourney, Adobe Firefly, and open-source tools like Stable Diffusion, Flux, and DALL-E 3 come back undetected. It also can’t verify that a real photo is authentic, since it only searches for its own watermark and not the absence of one. In other words, a clean result doesn’t mean the image is real.

So where does that leave us? Until detection goes universal or the industry agrees on one labeling standard, your best defense is still your own eyes: check the hands, the text, the shadows, and the sources. The tool is a step forward, just a small one.

Would you actually use a detector like this, or do you trust your own gut more? Drop your answer in the comments.