Google introduces 'AI Mode' for agentic web search experience: How it works

At I/O 2025, Google showcased how its 'AI Mode' in web search automate complex search tasks and streamline user workflows through agentic browsing capabilities

At the keynote session of Google I/O 2025, Google shared details on the evolution of its AI Mode, designed to deliver an agentic web search experience. Initially introduced in March, AI Mode has since been enhanced to offer users faster, more intuitive, and more comprehensive search results. Here are the new updates coming to the AI Mode:

AI Mode: Automating complex search tasks

Google said that the AI Mode represents a significant shift in how web searches are conducted. Instead of requiring users to sift through multiple pages of results, the tool autonomously handles the more tedious aspects of online research.

 

For complex topics – where standard search methods may prove insufficient – AI Mode acts as a digital agent. It uses a technique called query fan-out, which dissects the user's primary query into smaller, more specific sub-queries. It then conducts multiple searches in parallel, collecting information from a wide range of sources. The system compiles this data into a single, in-depth response, complete with source links for verification.

 

According to Google, this entire process is completed within seconds, significantly reducing the time and effort typically required to gather relevant and trustworthy information.

 

AI Mode: Agentic web search rollout

The rollout will initially kick off in the US and support tasks such as buying event tickets, making restaurant reservations, and booking local appointments. Google is partnering with platforms including Ticketmaster, StubHub, Resy, and Vagaro to deliver a seamless, integrated user experience.

 

What Is Project Mariner

Project Mariner is an experimental initiative built on Google’s Gemini 2.0 AI model, aiming to redefine human-computer interaction within web browsers.

 

According to Google, Project Mariner is designed to understand and interpret on-screen content, including text, images, code, and form elements. It enables the AI to interact meaningfully with websites by processing both visual and structural components, allowing it to complete complex tasks such as filling out forms, navigating UI elements, and compiling personalised search results.