From Query Box to Agentic Hub
Google Search is shifting from a static results page to an agentic hub, with Gemini 3.5 Flash now powering AI Mode for users globally. This latest Flash model is tuned for “sustained frontier performance for agents and coding,” making it the backbone of Google’s new AI search automation strategy. Instead of treating Search as a one-off question-and-answer tool, Google is reimagining it as a long-running assistant that remembers context and executes tasks over time. The multimodal search bar expands to handle conversational prompts and accepts text, images, videos, files, and even Chrome tabs as input. Once a query is submitted, Gemini-driven Google Search agents can continue working in the background, tracking information and surfacing updates without repeated manual checks. This multimodal search bar is effectively becoming a command center where discovery, analysis, and action all happen within a single, persistent interface.
Information Agents and Persistent Task Monitoring
At the heart of Google’s new agentic search features are information agents that operate continuously in the background. These Google Search agents watch the open web—blogs, news sites, social posts—and fuse that with real-time data across finance, shopping, and sports. Users can “brain dump” detailed requirements, such as specific apartment criteria or conditions for tracking stock movements, and let the agent monitor for changes. When a relevant listing appears or a trigger is hit—like a favorite athlete announcing a sneaker collaboration—the agent sends a notification instead of requiring users to re-run searches. This persistent model draws a line between traditional search and ongoing task management. Rather than a user repeatedly asking, “What’s new?” Google Search agents treat each query as the start of a long-running project. The system transforms search from passive information retrieval into a proactive, always-on monitoring service that anticipates user needs and responds in real time.
Agentic Coding and Real-Time Mini Apps in Search
Gemini 3.5 Flash also powers agentic coding in Google Search, enabling the platform to generate custom interfaces and tools inside the results page. Through generative UI capabilities, Search can assemble interactive visuals, tables, graphs, and simulations on the fly, tailoring layouts to match complex queries. Beyond these auto-generated experiences, users can craft their own persistent mini apps using Google’s Antigravity system. For recurring tasks—like planning a wedding, managing a home move, or maintaining a health routine—Search can build dashboards and trackers that pull in live reviews, maps, weather, and other real-time signals. A fitness tracker, for example, can combine local conditions, schedules, and location data into a single, continuously updated view. These mini apps live directly within Search, blurring the line between browsing and productivity software. Over time, the search results page becomes a workspace where users return to make progress, not just a destination for one-off answers.
Multimodal Command Center and Personal Intelligence
The new multimodal search bar is designed to unify discovery, conversation, and execution into a single workflow. Users can begin with an AI Overview, then seamlessly continue in AI Mode without thinking about which interface to choose. Google positions this as a frictionless continuum: ask a question, refine it conversationally, then let agents handle follow-up tasks in the background. Personal Intelligence, now expanded across nearly 200 territories and 98 languages, further personalizes these experiences by drawing insights from the user’s own context. Meanwhile, features like Universal Cart show how agentic reasoning extends into commerce: a single cart tracks products across retailers, hunts for deals, monitors price history, and anticipates complementary items. Together, these elements turn Search into a multimodal command center—an environment where AI search automation doesn’t just fetch links but orchestrates ongoing tasks, manages personal workflows, and proactively surfaces what matters next.
