Google marked the 25th anniversary of Google Images on July 14, 2026, by moving image generation directly into AI Overviews in Search and rolling out a redesigned home for the image product, extending a two-decade shift from text retrieval toward visual and generative search.

The announcement, published on the company's blog by Brad Kellett, Senior Engineering Director for Search, tied two new features to the milestone. One brings a browseable gallery to Google Images on desktop. The other places image generation inside AI Overviews, the summary panels that now sit atop a large share of search results. Both features were framed as extensions of a visual search history that began in July 2001.

The timing matters beyond the anniversary. AI Overviews has become one of the most contested surfaces in digital media, with publishers and independent researchers documenting sharp declines in the clicks that reach websites when the panels appear. Adding a generative image tool to that same surface deepens the reasons users have to stay inside Search rather than travel outward to the open web.

What Google shipped

Two updates accompanied the anniversary post, each on a staged rollout rather than an immediate global switch.

The first is a redesigned home for Google Images on desktop. According to Google, the new experience presents a dynamic gallery of images drawn from across the web, updated in real time and tailored to a user's interests. As a person browses and saves ideas to collections, those collections appear as tabs above the main gallery. Google said this would roll out over the following weeks on desktop in the United States, in English, for signed-in accounts.

The second update is the more consequential one for the search ecosystem. Google is bringing image generation directly into AI Overviews, powered by its Nano Banana model. According to the company, the feature turns a text prompt into a custom visual generated from scratch. Google said image generation in AI Overviews would begin rolling out over the coming weeks in English, across all regions that currently support image creation in AI Mode.

Elizabeth Reid, Vice President of Search, described the second feature in a post marking the anniversary. According to Reid, users will soon be able to ask Google to create a visual on the fly using the Nano Banana model, directly within search results, in cases where the exact image needed to explain an idea does not yet exist.

Neither the blog post nor the accompanying materials attached a specific date to the completion of either rollout, and both used the phrasing of a phased release measured in weeks.

The Nano Banana model behind the feature

Nano Banana is not new, and its scale is documented. The model, built on Gemini 2.5 Flash technology, first appeared in the Gemini app in August 2025. On October 13, 2025, Nano Banana expanded to Google Search and NotebookLM after users had generated more than 5 billion images since the initial launch, according to Naina Raisinghani, Product Manager at Google. That distribution shift moved the model beyond the single application it had occupied for roughly two months.

A more advanced version followed. On November 20, 2025, Google released Nano Banana Pro, built on the Gemini 3 Pro foundation, adding multilingual text rendering and search grounding for advertisers and creators. According to Raisinghani, the Pro system uses Gemini's reasoning and real-world knowledge to visualize information. The original Nano Banana continued to serve free-tier users after quota limits, while paid tiers gained access to the more capable model.

The AI Overviews integration announced on July 14 references the Nano Banana model rather than specifying the Pro variant in the anniversary post. Access to image generation inside AI Overviews is tied to the regions where image creation already functions in AI Mode, a scoping condition that limits the initial footprint to markets already cleared for generative visuals.

The anniversary post used the occasion to trace the product line's evolution, and the dates form a clear arc from retrieval toward generation.

Google Images launched in July 2001. According to Google, the catalyst was the 2000 appearance of a green Versace dress worn by Jennifer Lopez, which drove a volume of interest that a page of blue text links could not satisfy. The product made it possible to search and browse visual content from across the web for the first time.

Subsequent milestones layered new capabilities. Similar Images arrived in 2009, letting users find pictures by clicking a reference image rather than retyping a query. Search by Image followed in 2011, allowing an uploaded image or a pasted URL to function as the search term itself. Google Lens debuted in Search in 2018, turning a smartphone camera into a query tool for identifying objects, translating text, and surfacing product links.

Multimodal search advanced through the early 2020s. Multisearch in Lens arrived in 2022, combining an image and text in a single query. Circle to Search launched in 2024 as an Android gesture for selecting on-screen content; according to Google, it is available on more than 580 million Android devices. In 2025, Google combined Lens with AI Mode, introducing a technique the company calls "visual image fan-out," which breaks one image search into dozens of sub-queries to interpret a full scene.

The 2025 additions continued along that line. Search Live added a live camera feed paired with a voice conversation in AI Mode. Visual Results in AI Mode introduced conversational, shoppable image grids. Into 2026, Circle to Search gained multi-object recognition, again using the visual image fan-out technique, and an intelligent search box added the ability to ask detailed questions about uploaded images and receive an AI Mode response.

Why the AI Overviews surface is contested

The feature lands on a surface that sits at the center of an ongoing dispute over search traffic. AI Overviews summarize results at the top of the page, and a growing body of research has attempted to measure what that does to the clicks reaching publishers.

The measured declines are consistent in direction even as their magnitude varies by study. In April 2025, Ahrefs found that the first organic link loses an average of 34.5% of its clicks when an AI Overview appears, based on an analysis of 300,000 keywords. By February 2026, the same firm's follow-up work put the figure higher; Ahrefs recorded a 58% reduction in click-through rates for top-ranking pages, nearly double its earlier estimate.

Geographic data points in the same direction. According to SISTRIX figures from March 2026 cited in PPC Land's interview with Liz Reid, AI Overviews cut the position-one click-through rate in Germany from 27 percent to 11 percent, a shift the analysis associated with 265 million lost organic clicks per month across 100 million keywords.

A randomized study sharpened the debate. In August 2026, researchers found AI Overviews cut publisher clicks 39.8% in what the paper described as the first randomized design, comparing groups that saw the panels against groups for which the panels were suppressed. The study also reported no measurable difference in downstream engagement quality between the additional clicks surfaced when overviews were hidden and existing traffic, a finding that runs against Google's public characterization of AI Overview clicks.

Google has disputed that framing. Its executives have argued that AI Overviews absorb shallow clicks while sending higher-intent traffic onward, and Nick Fox, Vice President of Search, questioned the methodology of traffic-decline studies during a May 2025 podcast appearance. The two sides have largely traded correlations, and the randomized work was an attempt to break that stalemate.

Scale and adoption of the surrounding surfaces

The features tie into products that already operate at very large scale, which is part of what makes the integration significant for the marketing community.

Visual search volume is rising. According to Robby Stein, Vice President of Product for Google Search, visual searching on Google increased 65% year-over-year, with shopping queries dominating multimodal usage as people photograph clothing to shop the look or capture product images to find similar items. AI Mode itself had reached more than 200 countries and territories by October 2025, processing queries typically three times longer than traditional searches.

The commerce dimension is already established on these surfaces. Google Lens gained Shopping ads in October 2024, drawing on the Google Shopping Graph, a database the company described as covering more than 45 billion products; Google Lens introduced Shopping ads with existing AI-powered Search, Shopping, and Performance Max campaigns automatically eligible to appear. In May 2025, Google launched virtual try-on in AI Mode using personal photos, extending the shopping experience into apparel visualization.

Adding a generative image tool to AI Overviews sits alongside these commerce and discovery surfaces. It gives users a reason to produce a visual inside Search rather than seeking one elsewhere, which is consistent with the broader pattern of consolidation Google has pursued across its search and creative products.

What it signals for marketers and publishers

For the marketing community, the integration matters less as a single feature and more as a directional signal. Image generation moving into AI Overviews places creative production on the same surface where research has documented the steepest traffic losses for the open web.

The context is a broader routing of Google's capabilities through its Gemini model family and its owned surfaces. Nano Banana and Nano Banana Pro have already been folded into Google Ads through Asset Studio, into Search, into NotebookLM, and into the Gemini app, each expansion widening the range of tasks a user can complete without leaving Google's properties. The AI Overviews addition follows that logic on the organic side of Search.

Publishers face the surface from the other direction. AI Overviews reached more than 2.5 billion monthly active users by mid-2026, and an opt-out mechanism through Search Console leaves sites with a difficult trade, since withdrawing content forgoes traffic and ad impressions while withholding the AI-specific performance data needed to judge whether the withdrawal is worth it. A generative feature layered onto that surface does not change the traffic mechanics directly, but it reinforces the incentive for users to treat Search as a destination rather than a waypoint.

The anniversary framing casts the changes as continuity: a 25-year progression from finding a photo on the web toward creating one inside Search. The underlying shift is that the surface where the creation now happens is also the one whose effect on outbound traffic remains the central dispute in search economics.

Timeline

Summary

Who: Google, with the anniversary post authored by Brad Kellett, Senior Engineering Director for Search, and further remarks from Elizabeth Reid, Vice President of Search.

What: Google marked the 25th anniversary of Google Images by bringing image generation, powered by the Nano Banana model, into AI Overviews in Search, and by rolling out a redesigned, browseable home for Google Images on desktop.

When: The announcement was published on July 14, 2026, with both features described as staged rollouts over the following weeks.

Where: The Google Images redesign begins on desktop in the United States in English; image generation in AI Overviews rolls out in English across all regions that currently support image creation in AI Mode.

Why: The features extend a 25-year progression from text-based retrieval toward visual and generative search, and they place a creative tool on the AI Overviews surface that independent research has tied to significant declines in the clicks reaching publisher websites.