Google's AI Search box gets agents, coding, and Gemini 3.5 Flash
Google Search now runs background agents and lets you build apps directly in the box.
At I/O 2026, Google unveiled the biggest overhaul to its Search box in 25 years, replacing keyword-based queries with a conversational, multimodal interface. Users can now ask questions naturally (e.g., "I want a portable Bluetooth speaker for the pool, waterproof, with Alexa support") and include images, files, or videos. The AI Mode tab is globally powered by Gemini 3.5 Flash, which Google describes as a "major leap" for agents and coding, capable of reasoning across sources, handling longer prompts, and completing multistep workflows. AI Mode now serves over 1 billion monthly users, and the new Search box is rolling out immediately in regions where AI Mode is enabled.
Beyond the search box, Google introduced Search agents—persistent AI that work in the background to monitor topics, such as tracking price changes or news updates—and agentic coding, which allows users to build and run small applications directly within the Search interface. These features aim to transform Search from a simple link retriever into a personal assistant that can research, shop, book, monitor, and create on behalf of the user. The updates are part of Google's broader push to make Search more proactive and personalized, leveraging the Gemini 3.5 model family's speed and agent readiness.
- AI Mode now runs on Gemini 3.5 Flash, reaching 1 billion monthly users with faster, agent-ready reasoning across text, images, and video.
- New AI Search box accepts conversational, multimodal inputs (text, images, files, videos) and keeps context for follow-ups, rolling out globally.
- Search agents monitor background topics (e.g., price changes) and agentic coding lets users build apps without leaving Search.
Why It Matters
Google Search evolves from a link finder to a proactive assistant that researches, monitors, and builds for you in real time.