Google I/O 2026: Google Revamps Search Experience With AI Agents, Multi-Modal Queries

Google I/O 2026 was held at the Shoreline Amphitheatre in California on Tuesday. During the keynote of the annual developer conference, the tech giant announced a major overhaul of its Search experience, calling it the biggest change to the Search box in over 25 years. Among the most notable changes is the new intelligent Search box, built on Gemini 3.5 Flash. The revamped experience aims to make it easier for users to ask more natural, complex, and multi-modal queries using text, images, videos, files, and even Chrome tabs.

New AI-Powered Intelligent Search Box

In a blog post, Google said that the new intelligent Search box expands dynamically to provide users with more space to type detailed questions. It is claimed to be better designed to understand user intent, while also providing AI-powered suggestions that are more than autocomplete recommendations.

The tech giant has introduced multi-modal support, too. This allows users to search using text, images, files, videos, and Chrome tabs simultaneously. They will continue to receive traditional Search results alongside AI-generated responses.

At IO, Google also confirmed that Gemini 3.5 Flash is now the default model for AI Mode globally. The newer Flash model is claimed to offer improved reasoning, coding, and agentic capabilities while maintaining faster response times. Users can also continue asking follow-up questions directly from AI Overviews and transition into full AI Mode conversations across desktop and mobile devices.

AI Agents, Booking Features and Custom Dashboards

Google also introduced what it calls “Search agents”. As described by the tech giant, these are AI systems that are capable of continuously monitoring information and performing background research tasks for users.

The first implementation of this system will focus on “information agents,” which can monitor blogs, news sites, social platforms, finance data, shopping information, and sports updates based on user-defined objectives. Giving an example, Google explained that while apartment hunting or tracking sneaker releases, AI agents can continuously scan the web and notify users when the set conditions are met.

Information agents will initially be available in the summer for Google AI Pro and Google AI Ultra subscribers.

Agentic booking features within Search are also being expanded. This allows users to search for local services using natural language queries. They can initiate requests such as finding private karaoke rooms with food service availability, and Search can then aggregate pricing and availability information and redirect users to booking providers. AI will also be capable of calling businesses on behalf of users in select categories. These features are expected to roll out in the US later this summer.

Another major addition is agentic coding and AI-generated interfaces powered by Google Antigravity and Gemini 3.5 Flash. The company claims Search is now capable of building custom visual layouts, simulations, graphs, tables, and interactive interfaces tailored to user questions.

Lastly, Personal Intelligence in AI Mode is being expanded to more than 200 countries across 98 languages. Users can connect services such as Gmail and Google Photos, with Google Calendar support arriving later. The company emphasised that users retain full control over whether connected apps can be accessed by AI systems.