From Query Box to Gemini-Powered Control Surface
Google has rolled out what it calls the biggest overhaul of Search in 25 years by wiring Gemini 3.5 Flash directly into Google Search AI Mode. Flash is now the default model behind AI Mode globally, replacing older Gemini models and enabling faster, more capable responses across tasks like coding and complex reasoning. The classic search bar has been reimagined: it expands to support long, conversational prompts and accepts text, images, videos, files, and even live Chrome tabs as input. Users can ask a question, upload a screenshot, or reference an open tab in a single multimodal query. Under the hood, Gemini 3.5 Flash is tuned for sustained frontier performance, especially for agents and code generation, while remaining significantly faster than previous frontier models. The result is that Search begins to look less like a list of links and more like a responsive AI surface that can adapt to whatever the user is trying to do.
Information Agents: Search That Keeps Working After You Leave
The most transformative change is the arrival of multimodal search agents that operate as always-on information scouts. Google’s new information agents run in the background, continuously reasoning over the web and Google’s freshest data sources to track evolving topics on a user’s behalf. Instead of repeatedly searching for the same thing, you can describe your needs once—such as detailed apartment preferences, stock conditions, or niche product drops—and let the agent monitor blogs, news, social posts, finance feeds, shopping data, and sports updates. When something matches your criteria, it proactively alerts you, turning Google Search AI Mode into a persistent companion rather than a one-off query box. These agentic search features begin rolling out first to AI Pro and Ultra subscribers, and are designed to plug into Google’s broader agent ecosystem, where similar behind-the-scenes assistants also live inside the standalone Gemini experience.
Agentic Coding: Mini Apps and Dashboards Inside the Search Box
Gemini 3.5 Flash doesn’t just answer questions—it builds tools. Through what Google calls agentic coding in Search, users can now ask the search bar to generate custom interfaces on the fly, powered by the Antigravity development platform. At one level, Search can assemble generative UI components—tables, interactive visuals, graphs, or simulations—around a query, such as creating an interactive astrophysics visualization beside traditional results. At a deeper level, users can collaborate with Gemini 3.5 Flash to code bespoke mini apps directly in the results page. Think of recurring tasks like planning a wedding, managing a move, or designing a new health routine. Search can generate persistent dashboards or trackers that pull in live reviews, maps, weather, and other real-time data, then update continuously over time. These mini apps effectively turn the search bar into a programmable workspace, where every query can evolve into a persistent, agent-powered tool.
Persistent Agents Blur the Line Between Search and Gemini
As agentic features accumulate, the boundary between Google Search AI Mode and the Gemini app is starting to dissolve. Both environments now rely on Gemini 3.5 Flash as a shared engine, both support multimodal inputs, and both are gaining persistent agents that remember context and act over time. Personal Intelligence, which lets Gemini draw on data from services like Gmail or Google Photos for more contextual answers, is being expanded to nearly 200 countries and territories across dozens of languages. Meanwhile, Search has been tuned to remove friction between AI Overviews and AI Mode, flowing follow-up questions directly into the more conversational interface so users no longer need to decide where to continue. In practice, Search is now functioning as a front door to Google’s entire agentic platform. Whether you start in the search bar or the Gemini app, you are increasingly talking to the same underlying constellation of multimodal search agents.
