Over the past couple of years, Google has been injecting AI-driven features into Search. These efforts culminated in AI Mode, which debuted in 2025. Now, the tech giant has announced more upgrades to the search engine. Revealed during this year’s Google I/O keynote, the changes include what the company calls the “biggest upgrade” to its Search box in more than 25 years.
A core part of the update is Gemini 3.5 Flash, which will now serve as the default model in AI Mode for all users globally. This is the newest Flash model, offering sustained performance in terms of agentic capabilities and coding.
A revamped Search box
At base, Google’s search engine aims to help users ask just about anything, from simple queries to those that are more complicated and specific. That said, some people may find it difficult to condense their questions into short keywords. As such, the company is redesigning the Search box to better accommodate longer queries.
Rolling out to countries that support AI Mode, this new “intelligent” Search box will expand to give users more space to describe what they need. Additionally, it will offer suggestions that go beyond autocomplete. As the company puts it, the feature can provide nuances that the users “might not have even thought to add”. Other than text, users can also rely on a variety of inputs, including images, files, videos, and Chrome tabs.
In addition to this, Google is streamlining Search by allowing users to ask follow-up questions from an AI Overview. As of the official announcement, this new experience is available globally.
Search agents and generative UI
An upcoming feature is the ability to leverage information agents for updates. These agents can operate continuously in the background to gather data on specific queries and notify the user when they find the relevant details. For instance, someone looking for housing can specify their requirements and allow the agent to comb the web over time for listings. Once a match has been found, it will send a notification, which the user can then act on. Essentially, it’s like Google Alerts, but more tailored to the individual’s needs.
Furthermore, Search will be able to create interactive widgets to answer specific queries. This feature leverages Google Antigravity and Gemini 3.5 Flash to produce custom generative UI on the fly, including visual tools and simulations. In an example, a user can ask Search to explain how black holes affect spacetime, and the results will include an interactive visual demonstrating the concept.
This is not limited to one-off questions, either. According to Google, users will be able to create custom dashboards and trackers for ongoing tasks, sort of like “mini apps”. Use cases include designing a fitness tracker that suggests types of workouts and meals based on the person’s schedule and grocery lists.
As for when these features are coming, the company is aiming for a summer rollout. For now, information agents will be reserved for Google AI Pro and Ultra subscribers, while all users will be able to access the generative UI tools. That said, only paying subscribers in the US will be able to build custom mini apps. On that note, US users will also have access to more agentic booking capabilities.
(Source: Google [1], [2])
