Google Announces Search Tools To Distinguish AI-Generated Images

With how fast artificial intelligence is moving, there’ll come a time when you won’t be able to differentiate between photography and photorealism. It’s already happening as we speak; one of the highlights of 2023 was when images of the Pope in a puffer jacket made their rounds and fooled many on the internet.

To prevent more of such pictures from pulling a fast one on humans, Google is rolling out a couple of features to give users more background context about images that turn up in its search results, such as whether they are AI-created. The announcement comes hand-in-hand with new enhancements the tech giant plans to make for its search engine, including AI chatbot integration.

Google cites a 2022 Poynter study stating that 62% of people are convinced they are exposed to misinformation on a daily or weekly basis. AI’s growing prevalence will likely aggravate this problem, so the company is taking measures to inform users about potential mistruths whenever it can.

About This Image

Firstly, Google Search is introducing ‘About this image’, which works similarly to ‘About this result’ but provides insight into an image’s history when you click on the three dots assigned to it or upload the file onto Google Lens.

The new feature will show you when an image or similar-looking results was first indexed by Google, where it may have first surfaced online, and where else you can access it online. With that, the result could also be supplemented with relevant articles and fact-checking posts that may tell you if the image was, say, AI-generated.

Image via Google

“With this background information on an image, you can get a better understanding of whether an image is reliable—or if you need to take a second look,” explains the company in a new blog post.

In time, users will also be able to access ‘About this image’ anywhere on the web by right-clicking or long-pressing a suspicious image on the desktop and mobile versions of Google Chrome. The feature, however, will only have a limited initial launch in the US, and be restricted to English.

Markups for AI-generated images

In the same vein, Google is bringing in a new markup that lets artists and publishers tag images as AI-generated. The new label will show up on Google Search results, transparently letting visitors know if a “photo” is, in fact, synthetic.

Very importantly, these “AI-generated” markups will additionally be embedded in the original files, so anybody viewing those images from another website can identify their validity in the metadata itself.

Image via Google

This implementation will come with some weaknesses, though. Tagging is a manual process, which means it’s up to an image’s owner or publisher to disclose the authenticity of their work. According to Search Engine Land, Google admits its algorithm is “not there yet” when it comes to automatically detecting visuals as AI-made, so the responsibility is still on humans.

Anyhow, the company asserts that many publishers, like Midjourney and Shutterstock, are interested in adopting the feature.

When can I expect these features?

Google notes that ‘About the image’ and the labels for AI-generated images will debut “in the coming months,” though it hasn’t provided a specific date.

[via TechCrunch and Search Engine Land, images via Google]