The year was 2000, and the red carpet at the 42nd Annual Grammy Awards was about to witness a pop culture earthquake. Jennifer Lopez stepped out in a sheer, green silk chiffon Versace dress, held together largely by hope and a well-placed safety pin. It instantly became an iconic fashion moment, but behind the scenes, it also triggered a full-blown digital crisis.

Millions of people rushed to their computers, desperate to see what all the fuss was about. At the time, if you wanted to find something on Google, you typed in a query and received a neat, text-based list of blue links. But people didn’t want to read an essay about a dress; they wanted to look at it. The sheer volume of users trying to hunt down a photo of that green gown effectively broke the search engine’s capabilities, exposing a massive blind spot in how we navigated the early web.

Recognizing that the internet needed to evolve past the written word, a small team of engineers got to work. By July 2001, Google Images was officially born, launching with a catalog of roughly 25 million indexed pictures.

Fast forward to today, and that simple grid of pictures is celebrating its 25th anniversary. What started as a reactionary tool to help people look at celebrity dresses has quietly transformed into the foundational bedrock of how we interact with the physical world.

Over the last two and a half decades, the way we search visually has undergone a radical, AI-driven metamorphosis. In the early days, Google Images relied almost entirely on text—matching your query to the file names, alt-text, and surrounding captions of images hosted on websites. If a photo of a cat was mislabeled as a dog, Google’s algorithms were easily fooled.

The first major shift toward true visual understanding arrived around 2009 with the introduction of “Similar Images,” which allowed users to find related pictures based on visual cues rather than just text labels. By 2011, the search engine flipped the script entirely with “Search by Image,” letting users upload an actual photo to track down its original source or find visually similar copies across the web.

But the real magic happened when visual search migrated out of the browser tab and into our smartphone cameras. The launch of Google Lens in 2018 turned the lens into a live search box, giving users the ability to point their phones at an unfamiliar plant, a foreign menu, or a pair of shoes on the subway to get instant answers.

Today, that technology has become incredibly seamless. Features like Circle to Search allow users to highlight anything on their phone screens without even switching apps. More recently, the underlying tech has advanced to handle complex tasks like multi-object recognition and “visual image fan-out”—a technique where a single image search is broken down into dozens of sub-queries simultaneously, allowing you to deconstruct and shop an entire outfit from a single photo.

To mark this quarter-century milestone, Google is pushing the boundaries of visual search even further. The company has rolled out a brand-new, immersive home for Google Images, turning it into a dynamic gallery tailored to individual interests where users can organize ideas into dedicated collections.

Perhaps the biggest leap, however, is the integration of image generation directly into Search’s AI Overviews. Powered by Google’s latest Nano Banana model, the tool allows users to bridge the gap between finding what already exists and creating what doesn’t. If the perfect picture isn’t out there on the web, a simple text prompt can now generate a custom, high-quality visual right in the search bar.

From a frantic scramble for a Grammy dress to an AI-powered ecosystem that understands context, depth, and live video feeds, Google Images has fundamentally rewritten our relationship with technology. It proves that in a world increasingly dominated by screens, sometimes a thousand words still can’t compete with a single picture.