Google Rebuilds Search With AI Mode and Gemini 3.5 Flash
Trending • 8 hours ago • 6 min read
Updated Jun 3, 2026
For a quarter of a century, the act of searching the web looked the same. You typed a few words into a rectangle, and Google handed back a column of blue links to sort through yourself. That ritual, learned by billions of people, is now being dismantled. The biggest upgrade to the Google search box in more than 25 years began reaching users this spring, and the change runs deeper than a redesign. Increasingly, the page Google returns is not a list of places to go but a custom answer it has written for you on the spot, one that opens into a back-and-forth conversation if you want to keep going.
The Search Box Learns to Think
Unveiled at Google I/O 2026 on May 19 and now rolling out broadly, the centerpiece is a reimagined search box that the company describes as completely rebuilt around AI. Instead of a static field with simple autocomplete, it expands dynamically as you type, anticipates your intent with AI-powered suggestions, and accepts text, images, files, videos, and even Chrome tabs as input. The model doing the heavy lifting is Gemini 3.5 Flash, now the default in AI Mode globally. Google calls it a frontier-grade model that runs at Flash speeds, generating output several times faster than rival systems while costing a fraction as much to operate.
That efficiency is the whole point. "Part of the reason we focus on delivering frontier models, highly capable, but also very efficient, fast, and at a lower price, is because we want to bring it to as many people as possible," Sundar Pichai said in a press briefing ahead of the conference. The strategy is working by Google's own numbers. AI Mode passed one billion monthly users within a year of launch, and the company says queries to it are more than doubling every quarter.
Saying Goodbye to the 10 Blue Links
Elizabeth Reid, the head of Search, summed up the direction bluntly from the stage, describing the experience as "AI search through and through." For everyday users, that means fewer blue links and more answers generated directly on the results page, often assembled into a layout built specifically for the question. Through a feature Google calls Generative UI, Search can now construct interactive visuals, tables, charts, simulations, and small task-specific tools in real time, drawing on the agentic coding capabilities baked into Gemini 3.5 Flash. Ask for a workout plan and you may get a working tracker rather than a list of articles about workout plans. Those generative interfaces arrive for everyone in Search this summer, free of charge.
Reid was careful to add a caveat, telling the audience the new box does not mean you will only ever see AI responses. Classic results still exist, and Personal Intelligence features that connect Gmail, Google Photos, and soon Calendar are expanding to 98 languages across nearly 200 countries. But the default gravity of the product has shifted. The answer comes first, and the open web comes second.
Agents That Never Stop Searching
The most forward-looking addition turns Search from something you visit into something that works while you sleep. New information agents run continuously in the background, scanning blogs, news sites, social posts, and Google's freshest real-time data on finance, shopping, and sports to track whatever you care about. Hunting for an apartment in a tight market, or waiting on a limited sneaker drop, becomes a standing instruction rather than a daily chore. The agents notify you when something matches. They launch this summer for Google AI Pro and Ultra subscribers.
Google is pushing the agentic idea further still with expanded booking that can call businesses on your behalf to arrange home repair, beauty, or pet care appointments, also rolling out in the United States this summer. Taken together, the features sketch a Search product that increasingly does things instead of merely finding them.
An Extinction-Level Event for Publishers
For the websites that have spent two decades optimizing for those blue links, the math is alarming. When Google answers a question on its own page, the user has little reason to click through, and the traffic that funds independent journalism and niche expertise can evaporate. The early data is stark. HubSpot has reported losing 70 to 80 percent of its organic traffic, the education company Chegg disclosed a 49 percent decline, and the UK's DMG Media documented drops as steep as 89 percent on some queries. NPR has called the trend an "extinction-level event" for online news publishers.
Google's defense is that it still sends meaningful traffic and that the links remain available. Critics counter that defaults decide behavior, and that a generation of users learning to expect a finished answer will rarely scroll for sources. The deeper worry is structural. If the content that trains and feeds these answers stops being produced because nobody can fund it, the AI eventually runs out of fresh material to summarize. SEO professionals are already reframing their work around what some now call generative engine optimization, fighting to be cited inside the AI response rather than ranked beneath it.
The Race to Own the Answer
None of this is happening in a vacuum. The overhaul is Google's answer to a genuine threat. ChatGPT commands roughly 73 percent of the AI chatbot market, and OpenAI is moving fast to monetize it. Its advertising pilot, launched in February 2026, reportedly hit 100 million dollars in annualized revenue within weeks, with sponsored recommendations appearing at the bottom of responses. Perplexity has begun testing sponsored answers of its own. The contest is no longer about who has the best list of links. It is about where you go when you have a hard question, and every player wants that destination to be its own ad-supported surface.
Advertising is exactly where Google's overhaul gets delicate. Ads now appear in roughly a quarter of AI Mode results, up from around five percent in early 2025, and Google has signaled it will bring advertising into Gemini more broadly during 2026. The company is betting it can keep users inside an interface it monetizes while convincing them the answers are worth trusting. That balance, between a clean answer and a profitable one, will define whether AI search feels like an upgrade or a walled garden.
What is no longer in doubt is the direction of travel. The web Google indexed for 25 years was built for human eyes scanning a page of options. The web it is building now is built for a model that reads everything and replies in your voice. For users, that promises speed and convenience few will refuse. For the open web that made Google possible, the summer of 2026 may be remembered as the moment the search box stopped pointing outward and started answering for itself.
Sources
This article was researched using the following sources to ensure accuracy and reliability: