Google expands AI content checks across Search & Chrome

Google has expanded its content transparency and verification tools across Search, Gemini, Chrome, Pixel and Cloud, extending its effort to help users identify AI-generated and altered media.

The update brings together two main strands: broader use of SynthID, Google's digital watermarking system for AI-generated media, and wider support for C2PA Content Credentials, an industry standard that records how media was created and edited. Google says it has now watermarked more than 100 billion images and videos and 60,000 years of audio with SynthID.

Search and Chrome are being positioned as places where users can check the origin of online images, video and audio. In Search, people will be able to use tools including Lens, AI Mode and Circle to Search to ask whether an image was made with AI. Similar verification is also coming to Gemini in Chrome.

That verification feature has already launched in the Gemini app for image, video and audio and has been used 50 million times globally.

Camera records

Google is also expanding the use of Content Credentials on Pixel devices. Pixel 10 was the first smartphone to provide Content Credentials for images captured in its native camera app, and support for video on Pixel 8, 9 and 10 phones will follow in the coming weeks.

By adding provenance data at the point of capture, Google aims to distinguish original camera content from files later edited or generated with AI. Verification of C2PA Content Credentials is also rolling out in the Gemini app and will later come to Search and Chrome.

That will let users check not only whether a file carries an AI watermark, but also whether it is an unaltered camera original or has been modified, and by which tools. The approach reflects a broader market shift from simple AI labelling toward more detailed records of source and editing history.

Cloud launch

Alongside the consumer-facing changes, Google is launching an AI Content Detection API on Google Cloud's Gemini Enterprise Agent Platform. The product is aimed at organisations that need to identify synthetic media produced by Google models and other widely used systems.

Businesses could use the tool for internal tasks such as sorting feeds and preventing insurance fraud, as well as for public-facing uses including fact-checking and labelling synthetic media. Google is launching the API with a group of trusted partners and plans to refine it using their feedback.

The announcement highlights the commercial importance of provenance tools as generative AI becomes easier to use and harder to detect through visual inspection alone. For cloud providers, media authentication is becoming part of a broader push to sell AI governance and risk-management products to companies handling large volumes of digital content.

Industry backing

Google is also pushing for broader industry adoption of its watermarking technology. OpenAI, Kakao and ElevenLabs are bringing SynthID to more of their AI-generated content, extending the reach of one of the best-known watermarking systems on the market.

It has also open-sourced its SynthID text watermarking technology and is working with NVIDIA to watermark AI-generated video from the Cosmos world foundation models. The aim is to increase interoperability at a time when platforms, model makers and device manufacturers are developing different ways to signal whether media is authentic or synthetic.

Google remains active in the Coalition for Content Provenance and Authenticity, or C2PA, where it sits on the steering committee. That work is intended to ensure content transparency tools function across services rather than staying tied to a single company's products.

One example is Meta's plan to label camera-captured media with Content Credentials on Instagram. That would mean photos and videos shot natively on Pixel phones are recognised and labelled as authentic when shared on Instagram.

The broader backdrop is a fast-growing debate over how online platforms should label AI-generated material without creating false confidence in incomplete detection systems. Watermarking and provenance records can help, but their reach still depends on how consistently tools are applied across cameras, editing software, AI models and distribution platforms.

Google also pointed to existing efforts, including YouTube labels for AI-generated content and its work with trusted testers on Backstory, a project focused on making detection tools faster and more reliable. The latest rollout extends that effort from isolated labels to checks embedded in search, browsing, device cameras and cloud software.

"Content transparency is a complex challenge, but we'll keep developing ways to push the technology forward and set a high bar for the industry. Our goal is to empower you with the tools needed to determine the history of any content you encounter."