Google Search is now AI search. Here is What That Actually Means
Imagine opening a search engine and typing: "Keep me updated when my favourite athletes drop new sneakers."
For 25 years, that sentence would have meant nothing to Google. A search engine is a retrieval mechanism; it works around a simple logic - you give it words, it matches those words to pages across the web, ranks them by relevance and authority, and returns a list of results for you to click through. The relationship between user and search engine has always been transactional in the most literal sense: you arrive with a query, it responds with links, and the interaction ends. What you typed above is not a query; it contains no retrievable fact, no navigable destination, and no answerable question. It is an instruction to a system that was never designed to receive instructions, from a person who expected that system to keep watching the world on their behalf until the answer became relevant enough to surface.
But that changed on Tuesday at Google’s I/O keynote in Mountain View. Liz Reid, Google’s VP of Search and the person who has led the engineering, product and design of the world's most widely used information product for more than 20 years, offered that very sentence on the stage.
And she said it not as a vision of what Search might one day become but as a matter-of-fact illustration of what it already does. Behind her, Sundar Pichai had spent the opening of the keynote marking a decade since Google declared itself AI-first, telling the audience that the company now processes 19 billion AI tokens every single minute and counts 13 products with more than 1 billion users, before adding that Search now brings generative AI to "more people than any other product in the world." Then he handed the stage to Reid, who did not reach for product-roadmap language or technical qualifications.
She said six words - "Google Search is now AI Search." And the question that those six words open, for every publisher, every advertiser, every enterprise operator and every person who opened a browser tab this morning and typed something into that familiar white box, is not whether Google has changed. It is whether anyone noticed it happening, and what it means now that it has.
A billion people are already living inside the transition
The scale of what was announced on Tuesday only becomes fully legible when you understand that AI Mode in Search has already crossed 1 billion monthly active users, a number confirmed on stage by Reid. That figure is not a projection or an early-adopter metric.
It means AI Search is already the default way a significant portion of humanity interacts with the web, and that this transition happened not through a single launch event or a conscious decision by users to adopt a new product, but through the gradual accumulation of changes to an interface so familiar that most people stopped noticing what it was doing years ago.
The Intelligent Search Box, described by Google on Tuesday as the most significant update to the search interface in 25 years, no longer accepts only text. Users can now drag and drop videos, PDFs, audio files and open Chrome tabs directly into the query. The box expands dynamically to accommodate long, complex, conversational prompts that would have looked alien in the single-line interface Google maintained from 1998 until this week.
AI Overviews, which have doubled every quarter since last summer, now function as a continuous conversational thread rather than a static block of text, allowing users to ask follow-up questions and drill into sub-topics without submitting a new search. Search Agents can now act rather than answer, booking reservations, pulling data across multiple websites and executing multi-step workflows from a single prompt. The entire system runs on Gemini 3.5 Flash, which Pichai described on stage as capable of generating output at nearly four times the speed of competing frontier models.
Reid's summary of what Search now does — "whether you want to find it, buy it, or book it, Search will help you get it done" — is worth sitting with carefully because it describes something categorically different from what Search has always been. Finding was always its purpose - buying and booking are completions, transactions, destinations that used to live on other websites, owned by other companies, reached through the outbound click that Google's entire advertising model was built around generating.
When Reid says Search will help you get it done, she is describing a product whose relationship with the rest of the web has changed in ways that have significant consequences for every business that has ever depended on Google sending traffic somewhere.
What Reid told a different audience four weeks earlier
On April 23, 4 weeks before standing on the I/O stage and declaring the transition complete, Reid sat for an hour-long conversation with Bloomberg's Odd Lots podcast, hosted by Joe Weisenthal and Tracy Alloway. The audience was investors and economists rather than developers, which meant the questions were harder in a specific way: Weisenthal and Alloway were not asking about product features. They were asking about money, about what happens to advertising revenue when people stop clicking through to websites, and about whether Google is structurally capable of building the thing that threatens its own core business.
Reid's account in that interview was more considered than her keynote declaration. She argued that AI Overviews reduce what she called "bounce clicks," the fraction-of-a-second visits in which a user clicks to a page, grabs a single fact, and immediately returns to Search. Those visits, she said, were always low-value for both publishers and users. Deeper engagement, the 5-minute article reads, the research sessions, and the purchases that begin with extended browsing still generate outbound traffic. "If what you were going to go in and do is read an article for five minutes," she told Bloomberg, "you're still interested in reading that article for five minutes." Google tracks whether people return to Search more often as a key indicator of the product's health, she added, framing increased query frequency as evidence that AI Search is expanding the overall market rather than simply redistributing it away from publishers.
The independent data complicates that account considerably. The Reuters Institute Journalism and Technology Trends 2026 report, drawing on Chartbeat data tracking publisher traffic globally, found that Google search referral traffic to publishers dropped by roughly a third. Google Discover referrals fell 21% year-over-year across more than 2,500 publisher sites. Seer Interactive's analysis found that organic click-through rate for queries triggering AI Overviews fell from 1.76% in 2024 to 0.61% in 2025, a decline of 61% in a single year. These are not bounce-click numbers. They describe a broad structural reduction in outbound traffic that spans query types and publisher categories, rather than a surgical removal of low-value visits.
What makes this tension significant is not simply that Google and independent researchers disagree about the data, though they do. It is the 4-week gap between the Bloomberg interview and the keynote. In April, Reid was carefully contextualising a transition she described as ongoing and nuanced for an audience of investors concerned about advertising revenue.
On Tuesday, she declared it complete and permanent. The careful contextualisation and the plain declaration are not contradictory. They are sequential. She was describing the river changing course while reassuring people downstream that the water level was fine, and then she stood on a stage in front of the world and said the course had changed.
What the standing instruction actually reveals
The sneakers example is worth returning to because it illuminates something the product announcement language does not quite reach. A standing instruction to a search engine is not a feature. It is a different kind of relationship between a person and an information system. The user who types "keep me updated when my favourite athletes drop new sneakers" is not searching for something. They are delegating a monitoring task to a system and expecting it to hold that task indefinitely, continuously monitor the world on their behalf, and surface the result when it becomes relevant without being asked again. The user is not going to search but search, in effect, is coming to them.
That shift from a tool you use to a system that works for you has implications that extend well beyond consumer convenience. In the Bloomberg interview, Reid described how AI Mode has already changed the structure of queries themselves, revealing how deeply the transition has reached. In the keyword era, users learned over years of experience to compress their actual needs into the two or three words a search engine could reliably process.
Someone thinking about finding a restaurant for a group of five, including a vegan and two children, not expensive, in a specific neighbourhood on a specific evening, typed "restaurants New York." The gap between the real need and the submitted query was enormous, and it existed entirely because users had internalised the system's limitations as a condition of using it. AI Mode inverts that entirely. "They tell you the real problem," Reid said. "They don't take their need and translate it into what the computer understands. They try to give the computer their actual need and expect us to do the translation." The cognitive burden of query formulation that users had carried for 25 years has shifted to the machine.
The consequence is that query data has become qualitatively richer and more personally revealing than anything the keyword era produced. The user who asks AI Mode to find a restaurant for their specific group with their specific constraints has disclosed a demographic profile, a social situation, a budget range and a neighbourhood without being prompted for any of it. Multiplied across 1 billion monthly users typing queries that run, as Google has confirmed, two to three times longer than traditional keyword searches, the depth of signal that AI Search now processes is of a different order entirely from what Search has historically seen. This is not only a product transformation. It is a data transformation, and its commercial and competitive implications for every company operating in the information economy are still being mapped.
The experiment is over, everywhere, for everyone
Every previous AI Search feature has carried a qualifier: limited to specific markets, available to logged-in users, subject to change, rolling out gradually. Tuesday's announcement carried none of those qualifiers. AI Mode powered by Gemini 3.5 Flash is rolling out globally on desktop and mobile simultaneously as the default Search experience. Not US-first. Not English-first. Not experimental.
The cautious, gradual language that has accompanied every AI Search launch since 2023 was absent from Tuesday's stage entirely, because the product is no longer cautious or gradual.
That simultaneity matters more than it might initially appear. The 1 billion people who already use AI Mode monthly are not an early-adopter cohort that the rest of the world is catching up to at its own pace. They are the leading edge of a global default that is now being applied to everyone, everywhere, at once. The web traffic economy, the publishing business model, the SEO industry, the advertising infrastructure built on the assumption that Search generates outbound clicks — all of it now operates inside a product whose relationship with outbound traffic has structurally changed, globally and permanently, without a transition period or an opt-out.
Pichai put the underlying logic plainly. "It's been 10 years since we pivoted the company to be AI-first," he told the audience at the Shoreline Amphitheatre, and the weight of that sentence was in what it implied about arrival. The decade of reorganising teams, redirecting investment, defending the advertising business while rebuilding the product beneath it, absorbing the competitive pressure of ChatGPT and Copilot and every other challenger that has emerged since 2022, has produced a single outcome: the search engine that more than 4 billion people use to navigate the world is no longer a search engine in the sense that any of them learned to use it.
The curiosity the industry should be sitting with is not about what Google announced on Tuesday. It is about the realisation, arriving now for some and later for others, that the white box everyone opens a dozen times a day has already become something else entirely, and that the six words Reid chose to describe that transformation were the understatement of the year.
