Google Updates Images Turns 25 With Personalized Discovery and AI Image Generation

Google is turning Images into a personalized discovery feed and adding AI image generation directly to Search. The changes also revive an old question: how much of the image-search experience should happen without visiting the publisher?

Google Images began with Jennifer Lopez’s green Versace dress.

After millions of people searched for photos of the dress following the 2000 Grammy Awards, Google realized that blue links were not enough. Google Images launched in July 2001.

Google Images becomes a more personalized feed

Google is replacing the mostly empty Images homepage with a personalized gallery selected around a signed-in user’s interests.

Users will also see saved Collections displayed as tabs above the gallery, making it easier to return to topics such as travel, home design, shopping or fashion.

The new experience moves Google Images closer to Pinterest and Google Discover. Instead of waiting for a query, Google will begin recommending images before the user searches.

The redesign will initially roll out to signed-in desktop users in the United States using Google in English.

AI Overviews can now create images

Google is also adding image generation directly to AI Overviews.

When an existing image does not match a request, Search can create a custom visual from the user’s prompt. Google says the feature uses its Nano Banana image model and will roll out in regions where image creation is already supported in AI Mode.

This changes the basic purpose of image search. Google Images was built to find visual material published on the web. Google can now decide that no existing image is suitable and generate the answer itself.

The missing Getty Images chapter

Google’s anniversary history leaves out one of the most important changes to the product: the 2018 removal of the “View Image” button.

Getty Images filed an antitrust complaint with the European Commission in 2016, arguing that Google displayed large, high-resolution images in a way that reduced visits to the original source websites.

Google and Getty reached a licensing partnership in 2018. Soon afterward, Google removed the direct image button and began steering users toward the webpage hosting the image.

Google Search Liaison acknowledged that the change happened partly because of the Getty settlement.

The goal was to send more traffic back to publishers and provide better context around ownership, attribution and copyright.

Google is pulling the visual journey inward again

The contrast is difficult to ignore.

In 2018, Google changed Image Search to encourage users to visit publisher websites. In 2026, it is adding personalized browsing, saved Collections and AI-generated visuals that give users more reasons to remain inside Google.

Publishers may gain exposure from appearing in personalized galleries, but visibility does not automatically become traffic. Users may browse, save or generate images without ever visiting the source website.

Bing Images is still easier to use

This is also where Bing Images has a clear usability advantage.

Bing still provides a more direct route to the image itself. For users searching for a logo, screenshot, chart, flyer or product image, that can make the experience much faster.

Google often sends users to a webpage where the image may be buried in the article, hidden inside a carousel, loaded through JavaScript or surrounded by unrelated material. You will have a very hard time finding it, and have to suffer through mass Google ads and site paywalls to figure out that the image probably isn’t the same one or version Google is thumbnailing.

Bing makes it easier to inspect the image directly.

Direct access does not mean an image is free to reuse. Copyright and licensing restrictions still apply. But access and permission are separate issues.

The moral is, that most SEO’s still keep Bing Images on browser speed dial.

What publishers should watch

Search marketers should monitor image-search impressions, clicks and click-through rates in Google Search Console as the new homepage rolls out.

A rise in impressions without a similar rise in clicks could indicate that users are consuming or saving more visual content inside Google.

Original photography may also become more valuable as generated imagery expands. Product photos, event coverage, location images, screenshots, diagrams and first-party research provide specificity that generic AI images cannot easily replace.

Publishers should continue optimizing filenames, alternative text, captions, surrounding copy and image landing pages. The landing page should also make the promised image easy to find.

The larger question

Google Images began as a tool for finding pictures published by others.

Its next phase is broader. Google wants to recommend what users should see, organize what they save and create what cannot be found.

That makes the product more capable, but it also revives the issue behind the Getty dispute: how much value can Google extract from visual content before the publisher becomes optional?

Meanwhile, Bing continues to show that image search does not always need another layer of AI. Sometimes the best feature is simply a clean, direct link to the image the user wanted.