Google will label AI-generated and edited images to help you distinguish real from fake

Google is set to introduce new technology that will be able to identify whether an image was captured on camera, edited with software like Photoshop, or created with generative AI models. This feature, integrated into Google search results, will give users more transparency about the origin of images they come across online.

C2PA authentication standard

Google’s system is based on the Coalition for Content Provenance and Authenticity (C2PA) standard, an initiative backed by big tech companies like Amazon, Microsoft, Adobe, OpenAI, and Intel. C2PA creates a digital footprint for images by embedding information about their provenance in both hardware and software.

Google played a key role in developing the latest C2PA technical standard (version 2.1) and will use the upcoming C2PA trust list to verify the authenticity of image metadata. Laurie Richardson, vice president of trust and security at Google, told The Verge, “For example, if the data shows that a photo was taken with a specific camera model, the trust list helps confirm that this information is accurate.”

Integration with Google Search and other services

Google’s updated “about this image” feature in search results will clearly indicate whether an image was created or edited using AI tools. The company also plans to integrate C2PA metadata into its advertising systems to enforce AI-generated image policies.

“Our goal is to gradually scale this up and use C2PA signals to inform how we enforce key policies,” Richardson says. “We’re also considering ways to communicate C2PA information to YouTube viewers when content is captured on camera, and we’ll have more information on that later in the year.”

Challenges and opportunities

While Google’s adoption of the C2PA standard is a significant step toward addressing the proliferation of AI-generated images, widespread adoption and interoperability remain challenges. Currently, only a handful of Leica and Sony cameras support the C2PA standard, and broader support from camera manufacturers like Nikon, Canon, Apple, and Google is needed.

Software support is also limited, with Adobe Photoshop and Lightroom among the few applications capable of adding C2PA data. Encouraging broader adoption in software and online platforms is critical to the initiative’s success.

“Determining and signaling content provenance remains a complex challenge, with a range of considerations based on the product or service,” Richardson acknowledges. “And while we know there is no silver bullet solution for all online content, working with others in the industry is key to creating sustainable and interoperable solutions.”