Google Brings Gemini 3 AI Model to Search and AI Mode (blog.google)

Google has integrated Gemini 3 — its most advanced multimodal model — directly into Search’s AI Mode, available day one for Google AI Pro and Ultra subscribers in the U.S. (select “Thinking” from the model menu). Gemini 3 Pro brings stronger reasoning and intent understanding to Search, upgrading Google’s query fan-out technique so Search can run more targeted subqueries and surface relevant web content it might previously have missed. Google will also add automatic model selection to route harder AI Mode and AI Overview questions to Gemini 3 while keeping faster models for simpler tasks. Beyond text answers, Gemini 3’s deep multimodal and agentic coding capabilities power a new generative UI that creates dynamic visual layouts, interactive tools and real‑time simulations tailored to your query — for example, an interactive three-body physics simulator or a custom mortgage calculator embedded in the response. The change promises richer, more actionable search results and new pedagogical and productivity workflows, but also raises practical implications around compute use, latency, content credibility and evaluation of generated tools. For researchers and developers, Gemini 3 exemplifies progress in combining advanced reasoning, multimodal understanding and on-the-fly code generation to produce bespoke, interactive user experiences within search.