Google expands Gemini in Chrome with 'auto browse' in India: What's it

Google's auto browse in Chrome can handle routine web tasks such as booking parking, filling lengthy forms and ordering groceries and more

Google has expanded Gemini in Chrome to Android users in India, while also introducing its agentic auto browse feature in the country. The feature, which is coming to Gemini in Chrome on both desktop and Android, allows the AI to take actions across multiple websites and handle multi-step tasks on a user’s behalf. Gemini in Chrome was previously limited to desktops.

 

Gemini in Chrome comes to Android: What’s it

 

Gemini in Chrome, which was introduced on desktops in India earlier this year, is now available on Android as well. It allows users to ask questions about web pages, summarise long articles and connect Gemini with Google apps such as Keep and Calendar without leaving the browser.

 

 

The feature can also use Gemini’s image-generation capabilities, including Nano Banana, allowing users to create or modify images while browsing on their phones. 

Auto browse handles multi-step tasks

Auto browse is now available on desktop and Android in India for Google AI Pro and AI Ultra subscribers. Unlike conventional browser features such as autofill, which mainly enter information into forms, auto browse is designed to handle a sequence of actions. It can research information across websites, fill out forms and complete other repetitive online tasks.

 

Google mentioned examples including booking parking for an event, ordering groceries, researching hotel and flight prices across different dates and collecting information needed for everyday tasks. The feature is also designed to work with more complex workflows. For instance, users could ask it to find products based on an image, compare prices, add suitable items to a shopping cart and apply available discounts. 

User Control

Google said auto browse has been designed with safeguards because the feature can interact directly with websites. Its AI models are trained to detect known threats such as prompt injection, where malicious instructions embedded in a webpage attempt to manipulate an AI agent.

 

The system can also ask users for confirmation before completing certain sensitive actions. This is intended to keep the user involved when an action could have financial, privacy or other important consequences.