“From a box that returns links to an agent that books your dinner, calls your plumber, and monitors the news while you sleep — Google’s transformation of Search is the biggest shift in information retrieval in a generation”
For more than two decades, using Google Search meant typing a few words into a white box and choosing from a list of blue links. That ritual, familiar to billions of people, is being dismantled with a swiftness that has caught even veteran observers off guard. The engine powering this transformation is Gemini, Google’s flagship artificial intelligence system, and what it is doing to Search goes far beyond adding a chatbot to the top of the results page.
At Google I/O 2026 held earlier this month, the company announced that AI Mode in Search had crossed one billion monthly users, while AI Overviews, which place generated summaries above traditional results, now reach 2.5 billion people every month. Those numbers are staggering. To put them in context: for the first time in the internet’s history, a significant portion of the world’s population is receiving answers produced by a machine rather than selecting from sources curated by an algorithm. The difference is more than cosmetic. It is architectural.
The most consequential change is the introduction of what Google calls “information agents.” These are AI systems that run continuously in the background, monitoring news cycles, financial markets, shopping trends, and other topics on a user’s behalf. Rather than waiting to be asked a question, they surface summarised briefings and recommendations proactively. Google says information agents will launch initially for its AI Pro and Ultra subscribers this summer, a detail that signals something else worth noting: the most capable version of Search is becoming a premium product.
“Search is no longer a retrieval tool. It has become an assistant that acts, books, calls, and decides on your behalf.”
Alongside agents, Google unveiled what it described as the biggest redesign of its Search box in 25 years. Powered by Gemini 3.5 Flash, the new interface accepts text, images, video, audio, and even open browser tabs as input. The box expands dynamically as you type, encouraging longer, more conversational queries. For complex requests, Search now generates bespoke visual interfaces, mini-dashboards, and interactive displays rather than a static list of results. Google calls this capability Generative UI, built on its internal Antigravity platform, and it means that two users asking different questions may see entirely different formats in response.
The practical implications stretch well beyond how information looks on screen. Google is expanding what it terms “agentic capabilities” inside Search to include booking local services, reserving restaurant tables, and even placing phone calls to businesses on a user’s behalf. Someone searching for a private karaoke room for six people on a Friday night that serves late food can now ask Google to handle the entire booking, not merely surface a list of venues. For home repair, beauty, or pet care queries, Search can call the business directly. These capabilities are set to roll out across the United States this summer.
The underlying model driving much of this is Gemini 3.5 Flash, which Google describes as designed for coding, long-running tasks, and AI agents. It replaces the previous default model in AI Mode globally, and runs alongside Gemini Omni, a newly announced multimodal system capable of generating and editing video using text, images, and audio as inputs. Together, the two models form the technical backbone of what Google is calling a shift toward agentic AI, systems that do not merely answer questions but take action in the world on your behalf.
Critics and regulators will have no shortage of questions. If Search now generates the answer rather than directing you to the source, what happens to the publishers, journalists, and researchers who produce that underlying information? When an agent books a service, selects a vendor, or summarises a news story, who bears responsibility for errors? And as the most capable features are locked behind subscription tiers, does the democratising promise of the open web give way to a tiered information economy? Google has not offered complete answers to any of these questions, and the pace of rollout suggests it does not intend to wait until it does.
What is clear is that the transformation is already underway, at scale, and largely without fanfare. There was no single announcement, no dramatic unveiling that marked the end of Search as the world knew it. Instead, the change is arriving feature by feature, model by model, update by update. Google is not replacing Search. It is replacing what Search means.
