A new era for AI Search

TL;DR

• Google has updated its Search AI Mode with Gemini 3.5 Flash as the new default model globally, featuring an intelligent search box that supports text, images, files, videos, and Chrome tabs as inputs.
• AI Mode has surpassed one billion monthly users, with query volumes more than doubling every quarter since its launch.
• Search agents are now entering the market, starting with "information agents" that monitor the web 24/7 for specific triggers, launching for Google AI Pro and Ultra subscribers this summer.
• Agentic capabilities are expanding to include booking local services and making phone calls on behalf of users for home repair and care services in the U.S.
• Google Antigravity and Gemini 3.5 Flash enable "agentic coding" in Search, allowing users to generate custom UIs, visual tools, simulations, and persistent mini-apps on the fly.

Why It Matters

This shift redefines Search from a static retrieval tool into a dynamic, agentic platform that proactively monitors, books, and builds software for users. The integration of Generative UI and persistent mini-apps bridges the gap between querying information and executing complex, multi-step workflows within the search interface. For practitioners, this highlights the convergence of foundation models, agent frameworks, and user interfaces, signaling that the next competitive front in AI will likely be controlled by who can best manage agentic workflows at scale.

Key Data

• User Base: AI Mode has surpassed one billion monthly users.
• Growth Rate: AI Mode queries have more than doubled every quarter since launch.
• Model Specification: Gemini 3.5 Flash is the new default model for AI Mode globally.
• Rollout Scope: The new intelligent search box is available in all countries and languages where AI Mode is active.
• Geographic Availability: Agentic booking and coding features are launching in the U.S. first, with Personal Intelligence expanding to nearly 200 countries and 98 languages.

Technical Details

• Gemini 3.5 Flash is described as delivering sustained frontier performance specifically for agents and coding tasks, replacing the previous default in AI Mode.
• The new search interface dynamically expands to accommodate multi-modal inputs, including text, images, files, videos, and active Chrome tabs, allowing for context-rich queries.
• Information agents operate in the background to monitor diverse web sources like blogs, news, social media, and real-time data for finance, shopping, and sports, synthesizing updates and taking actions when criteria are met.
• Agentic coding utilizes Google Antigravity to build custom generative UIs in real-time, assembling interactive visuals, tables, graphs, and simulations without pre-built templates.
• The system supports persistent "mini apps" such as fitness trackers that tap into live data sources like reviews, maps, and weather to maintain functionality over time.

Industry Insight

• Search engines are evolving into agent orchestrators, where the value proposition shifts from answering questions to executing complex tasks like booking services or monitoring markets continuously.
• The introduction of generative UI suggests a future where software interfaces are ephemeral and task-specific, reducing the need for distinct applications for niche tasks.
• The tiered release strategy for agentic features (Pro/Ultra first) indicates that high-compute agentic capabilities may remain a key differentiator for premium subscriptions in the near term.

FAQ

Q: When will search agents be available to the general public?
A: Information agents and agentic booking capabilities are launching for Google AI Pro and Ultra subscribers first, with broader U.S. availability expected later this summer.

Q: Can Search build and maintain custom apps for me?
A: Yes, using Agentic Coding, Search can build custom "mini apps" like trackers or dashboards that tap into real-time data sources for ongoing use, starting with premium subscribers in the U.S. in coming months.

Q: How does the new search box handle complex inputs?
A: The reimagined search box dynamically expands to allow users to describe needs in detail and accepts multi-modal inputs including text, images, files, videos, and Chrome tabs to anticipate intent.