SynthID is Google DeepMind’s technology for marking content made by artificial intelligence. It embeds watermarks that people cannot see or hear.

The system tweaks AI-generated text, images, audio or video in tiny ways. The changes are invisible to humans but can be detected by software, even after editing like cropping or compression. For text, the system adjusts word choices in subtle patterns; for images and video, it changes individual pixels. Newsrooms can check suspect images before publishing, and social platforms can label flagged posts automatically.

The goal is trust. As AI tools produce realistic fake photos, voices and videos, watermarks help platforms and fact-checkers trace where content came from.

Google is building SynthID into its AI products. Other firms use similar methods, and regulators are pushing for clearer labels on AI-generated media. Watermarks are not foolproof, but experts call them an important layer of defence alongside media literacy and platform rules.

Context: watermarking is one of several defences against deepfakes and AI-driven misinformation.