Google has opened its SynthID Detector to users worldwide, offering a way to check whether images, videos, audio or text were made with its AI tools. But how does it actually work?

SynthID embeds an invisible watermark at the moment of creation. For images, tiny pixel adjustments encode a hidden pattern the eye cannot see. For text, the AI subtly shifts its word choices to form a detectable statistical signature.

To check a file, you upload it to the Detector portal. It scans the content and highlights the parts most likely to carry the watermark. The mark survives cropping, compression and filters because it lives in the content itself, not in metadata.

Google says over 100 billion images and videos now carry SynthID marks. The tool is not foolproof, but it is a promising weapon against deepfakes and AI misinformation.

Critics note the detector only works on Google-made content, so AI fakes from other companies slip through. Google says it is working with partners to widen coverage.