Google Search Gets Gemini 3

Here is the rewritten version, expanded into 15 clear, simple British-English paragraphs, made longer, natural, and not AI-sounding:

Google has introduced Gemini 3 Pro to the AI Mode in Google Search, marking a noticeable step forward in how the search engine processes and presents information. This update brings stronger reasoning skills, improved query handling, and brand-new visual layouts that adjust themselves to the user’s needs in real time.

For now, Gemini 3 Pro is available to Google AI Pro and Ultra subscribers in the United States. What makes this launch stand out is that it’s the first time Google has added a new Gemini model to Search on the very day of its release. It shows how quickly Google wants people to experience the model’s upgraded abilities.

Subscribers who want to access Gemini 3 Pro can do so by selecting the “Thinking” option within the model menu in AI Mode. This setting allows users to tap into the model’s deeper reasoning skills, which Google says can handle more involved and detailed questions than before.

Robby Stein, Google’s Vice President and General Manager of Search, announced the update on X. He described Gemini 3 as Google’s most intelligent model so far and expressed his excitement about making it part of Search immediately. This reflects Google’s broader aim of turning Search into a more interactive, helpful tool rather than simply listing links.

Google has also confirmed that Gemini 3 will be made available to all users in the U.S. soon, not just subscribers. However, those with Pro and Ultra plans will continue to enjoy higher usage allowances, meaning they can test more of the model’s advanced capabilities without running into limits.

According to Google, Gemini 3 brings state-of-the-art reasoning and a deeper understanding of multimodal inputs, meaning it can analyse text, visuals, and complex ideas together. This makes it suitable for handling more demanding queries and offering clearer explanations for difficult subjects.

One of the biggest changes is how AI Mode decides what a “helpful response” should look like. Instead of simply providing written text, the system can now build an entire layout around the query. This includes choosing when to add images, tables, diagrams, and other elements that make the answer easier to understand.

Robby Stein explained that Gemini 3 effectively reshapes the way AI Mode responds by generating layouts on the spot. The system looks at what the user is asking and designs the response from scratch, selecting the most suitable format for the information it needs to present.

This dynamic layout generation means the user may no longer see identical response styles for every question. Instead, the model adjusts itself depending on the complexity of the request. For example, a maths question might produce a table, while a visual problem may show diagrams or images directly in the response.

Another important update is the introduction of automatic model selection, which Google plans to release soon. With this feature, tougher or more technical questions will be routed to the most advanced model available, while quicker, simpler tasks will be handled by lighter versions for faster responses.

Gemini 3 also enhances Google’s query fan-out technique. This is the process where the search engine runs several related searches at the same time to make sure it fully understands what the user wants. With the upgrade, Search can now run more of these background searches and interpret user intent more accurately.

Because of this improved fan-out system, Google suggests that it may now uncover pages it previously overlooked. This could lead to more detailed and relevant results, especially for queries that require information from multiple sources or angles.

Generative UI features are another major addition. With these tools, Gemini 3 can build visual layouts specifically tailored to the question. This includes images, grids, tables, and even small interactive tools or apps created in real time to help users compare options or test scenarios.

Google shared examples such as an interactive physics simulation for understanding the three-body problem and a personalised mortgage calculator that shows long-term savings. These tools are built instantly and appear directly within the AI response, with clear links to reputable sources provided underneath.

As these updates roll out more widely, it may influence how website content is discovered and displayed in AI Mode. With more advanced query handling and new on-screen tools, the way links, citations, and source references appear could shift, especially during complex searches.

Google says it will continue refining these features in the coming weeks, using feedback from early users to improve the system further. Wider access to Gemini 3 in AI Mode is planned, though the exact timeline for full availability has not yet been announced.