Google Search Central updated its documentation to reflect support for labeling images that have been augmented or manipulated with AI. Google also quietly removed the “AI Generated” metadata from beta status, indicating that the “AI Generated” label is now fully supported in search.
IPTC Photo Metadata
The International Press Telecommunications Council (IPTC) is a standards-setting body that, among other things, creates standards for image metadata. Photo metadata enables an image to be labeled with information about the image, such as copyright, licensing and image description information.
Although the standards are created by an international press standards organization, the metadata standards they develop are used by Google Images in contexts outside of Google News. Metadata allows Google Images to display additional information about an image.
Google's documentation explains the use case and benefits of metadata:
“When you specify image metadata, Google Images can display more details about the image, such as who the creator is, how people can use the image, and credit information. For example, Providing licensing information can make the image eligible for a licensable badge, which provides a link to the license and further details about how someone can use the image.”
AI image manipulation metadata
Google quietly adopted metadata standards for images that were manipulated with AI algorithms commonly used to manipulate images, such as convolutional neural networks (CNNs) and generative adversarials. networks (GANs).
There are two forms of AI image manipulation covered in the new metadata:
Painting
Painting is generally conceived as filling in missing parts with the aim of restoring or reconstructing an image. But painting is also an algorithmic manipulation that adds to the image.
Out painting
Outpainting is the algorithmic process of adding to an image, extending it beyond the borders of the original image, adding more to it than what was in the original image.
Google now supports labeling images that combine both of these methods with a new metadata property of the digital source type called CompositeWithTrainedAlgorithmicMedia.
CompositeWithTrainedAlgorithmicMedia
Although the new property looks like structured data, it is not schema-structured data. This is the metadata that is embedded in a digital image.
This is what was added to Google's documentation:
“Digital Source Type
compositeWithTrainedAlgorithmicMedia: An image is a composite of trained algorithmic media with some other media, such as painting or outpainting operations.”
Label for “AI Generated” – Algorithmic Media Metadata
Google also lifted the beta status of the algorithmic media metadata specification, which means that images created with AI can now be labeled as AI generated if algorithmic media metadata is embedded within an image.
This is the document before the change:
“algorithmicMedia: The image was created purely by an algorithm that was not based on any sample training data (eg, an image created by software using a mathematical formula).
Beta: Currently, this property is in beta and is only available for IPTC photo metadata. Adding this property qualifies your image for display with an AI-generated label, but you may not see the label in Google Images yet, as we're still actively developing it. have been.”
The change in the documentation was to remove the entirety of the second paragraph to remove any mention of beta status. Interestingly, this change does not appear in Google's changelog.
Google's Search Central documentation changelog noted:
“Supporting a new IPTC digital source type
What: Added compositeWithTrainedAlgorithmicMedia to IPTC photo metadata documentation.Why: Google can now extract Trendalgorithmic Media IPTC news codes with Composite.
Read Google's latest documentation:
Image Metadata in Google Images
Featured image by Shutterstock/Roman Samborski