What changed: two new ways to explore and create
Google Images turns 25 this week, and to mark the occasion, Google is rolling out two updates that signal a shift in how we interact with visual content in Search. The first is a new browseable homepage for Google Images, featuring a dynamic, immersive gallery of images from across the web, updated in real time and tailored to your interests. As you browse and save ideas to your collections, they appear as tabs above the main gallery, making it easy to jump back in and continue exploring. This rolls out over the coming weeks on desktop in the U.S. in English, and requires signing in to your Google Account.
The second update is arguably more significant for product builders: image generation is coming directly into AI Overviews in Search. Using Google’s latest Nano Banana model, a simple text prompt can now be transformed into a high-quality, custom visual made from scratch. This feature will roll out over the coming weeks in English, for all regions that currently support image creation in AI Mode.
In other words, Search is no longer just about finding images—it’s about creating them. When you have a specific vision that doesn’t exist on the web, AI can now draw it for you, right inside the search results page.
How we got here: 25 years of visual search milestones
Google’s Senior Engineering Director for Search, Brad Kellett, traces the origin of Google Images back to 2000, when Jennifer Lopez’s iconic green Versace dress broke the internet. The standard search page at the time—a list of blue text links—simply wasn’t enough. People didn’t just want to read about the dress; they wanted to see it. So in July 2001, Google Images launched, making visual content searchable for the first time.
Since then, the evolution has been steady and revealing:
- 2009: Similar Images – Let you find pictures without relying on text alone. Search for “bow” and you might see hair bows and bows and arrows; clicking “find similar images” refined results without typing a new query.
- 2011: Search by Image – Uploading an image or pasting its URL turned visuals into search terms, letting you identify sources and find visually similar content.
- 2018: Google Lens in Search – Turned your smartphone camera into a search box, letting you identify objects, translate text, and pull up product links in real time.
- 2022: Multisearch in Lens – Marked a step into multimodal search, letting you combine text and images in one query, like snapping a photo of a landmark and asking “what inspired this design?”
- 2024: Circle to Search – Let Android users circle, highlight, scribble, or tap anything on their screen to search without switching apps. It’s now available on more than 580 million Android devices.
- 2025: Lens + AI Mode – Combined Lens’s multimodal power with AI Mode’s reasoning, using a “visual image fan-out” technique that breaks a single image search into dozens of sub-queries to understand full context.
- 2025: Search Live – Let you share your phone’s live camera feed while having an interactive voice conversation in AI Mode, capturing motion and surroundings.
- 2025: Visual Results in AI Mode – Introduced a new way to explore, imagine, and shop, letting you describe something conversationally (e.g., “barrel jeans that aren’t too baggy”) to get a grid of visual inspiration and shoppable products.
- 2026: Circle to Search Multi-Object Recognition – Expanded Circle to Search to explore multiple objects in a single image at once, again using visual image fan-out.
- 2026: Intelligent Search Box – The biggest upgrade to the Search box in over 25 years, letting you upload images (even multiple at a time) and ask detailed questions about them, with AI Mode responses.
This timeline shows a clear trajectory: from text to images, from static to dynamic, and from retrieval to generation.
Practical implications for product builders
For teams building AI products, these updates offer several concrete takeaways:
Input methods are expanding rapidly. Search inputs have evolved from text to images, camera, circling, live video, and now generative prompts. If your product relies on a single input modality, consider how you might support more natural, multimodal interactions. The trend is toward meeting users where they are—whether that’s a photo, a screen gesture, or a live camera feed.
Search and creation are converging. By embedding image generation directly into AI Overviews, Google is positioning Search as a content production tool, not just an information retrieval system. This could have profound implications for products that depend on search traffic: if users can generate the image they need instead of clicking through to a website, the nature of that traffic may shift. Think about how your product might offer value beyond being a source of pre-existing content.
Personalization and real-time updates are becoming baseline expectations. The new browseable homepage emphasizes real-time updates and intelligent tailoring to individual interests. Visual exploration is becoming more like a dynamic gallery than a static list of results. If your product deals with visual content, consider how you can make discovery feel more alive and personalized.
The “visual image fan-out” technique is a powerful pattern. Google’s approach of breaking a single image into dozens of sub-queries to understand context is a useful mental model for any AI system that needs to reason about complex visual scenes. It’s not just about recognizing objects; it’s about understanding relationships and intent.
Limitations and trade-offs to keep in mind
As with any new feature, there are important caveats:
- Geographic and language restrictions: Both new features are rolling out in English, and image generation is limited to regions that already support image creation in AI Mode. The browseable homepage is U.S.-only for now. Non-English markets and enterprise users may need to wait.
- Account requirement: The new homepage requires signing in to a Google Account, which may limit its reach in contexts where users are logged out or privacy-conscious.
- Generative AI is experimental: Google notes that the audio version of the blog post is “generated by Google AI” and that “Generative AI is experimental.” This applies to the broader AI features as well—expect occasional inaccuracies or unexpected outputs.
- Dependence on the Nano Banana model: The quality of image generation depends on the underlying model, which may have its own limitations in terms of style, accuracy, and safety filters.
For product builders, these limitations mean you should not assume immediate global availability. Plan for gradual rollout and consider how your product might work in regions or languages where these features aren’t yet available.
A concrete takeaway: design for the “see” moment
Looking at 25 years of visual search, one pattern stands out: the moments when users need to see rather than read are where the biggest innovations happen. Google Images was born from the desire to see Jennifer Lopez’s dress. Similar Images addressed the ambiguity of visual language. Lens made the physical world searchable. And now, image generation in AI Overviews addresses the gap when the perfect image doesn’t exist yet.
If you’re thinking about integrating visual AI into your product, start by asking: “When does my user need to see something that they can’t easily describe in words, or that doesn’t exist yet?” That’s the sweet spot where visual search and generation can add real value.
For example, an e-commerce app could let users generate a custom image of a product in a specific color or setting before buying. A design tool could let users describe a visual concept and get a starting point for iteration. A travel app could generate images of a destination at different times of day or seasons. The key is to identify the “see” moment in your user’s journey and make it as frictionless as possible.
Google’s 25-year evolution offers a roadmap: start with making visual content discoverable, then make it interactive, then make it generative. The next step for your product might be closer than you think.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.