Google is expanding its AI media verification tools across more of its products, bringing SynthID watermark checks and C2PA Content Credentials closer to everyday users in Search, Chrome and Gemini.
SynthID is Google DeepMind’s watermarking system for AI-generated media. It embeds signals into generated images, video, audio or text that are meant to survive common edits such as cropping, resizing, compression or filters. Google says SynthID has now been used to watermark more than 100 billion images and videos, plus 60,000 years of audio.
The important limitation is that SynthID is not a general-purpose lie detector for the internet. It can help identify content carrying a SynthID watermark, especially material created or processed through systems that support it. It does not prove that unmarked content is authentic, and it does not automatically catch every synthetic file made with another tool.
Google is also adding support for C2PA Content Credentials, the open provenance standard backed by the Coalition for Content Provenance and Authenticity. Content Credentials work differently from SynthID. Instead of an invisible watermark embedded into generated media, they attach signed metadata about how a file was created or edited, where supported. In plain English, SynthID is closer to a hidden mark; C2PA is closer to a tamper-evident label.
According to Google and reports from The Verge, C2PA verification is rolling out first in the Gemini app, with Search and Chrome support expected to follow. Google also plans to add Content Credentials to video captured on Pixel 8, 9 and 10 phones, expanding provenance from AI-generated material into ordinary camera capture.
For studios, broadcasters and platforms, the shift matters because provenance is becoming less of a specialist compliance tool and more of a consumer-facing expectation. Viewers, journalists, moderators, rights teams and marketing departments may increasingly ask not just whether something looks real, but whether there is a verifiable trail behind it.
The caveat is obvious and important. Watermarks can be absent, stripped, damaged, unsupported or disputed, and metadata systems only work when creation tools, platforms and distributors preserve the chain. Google’s own SynthID has also faced claims of reverse-engineering, which Google disputes. So this is not the end of synthetic-media uncertainty. It is another layer in the awkward stack of trust.
Still, the direction is meaningful. If AI media verification becomes available from a right-click in Chrome or a check inside Search and Gemini, provenance stops being an abstract standards conversation and becomes something audiences may expect to see. That has consequences for anyone releasing, licensing, archiving or authenticating media in public.