Home | My Feed | Saved Stories | About | Privacy
Back

Google Launches 'Auto Browse' AI Feature for Chrome, Initially for Subscribers

Show me the source 4 Sources
Generated on: Last updated:

Google Introduces "Auto Browse": AI Agent for Chrome Automates Online Tasks

Google has launched "Auto Browse," a new AI-powered feature for its Chrome desktop browser, designed to automate various online tasks. Powered by the Gemini 3 generative AI model, this feature functions as an AI agent capable of interacting with web pages to complete multi-step workflows. The initial rollout is limited to subscribers of Google’s AI Pro and AI Ultra plans in the United States.

Overview and Functionality

The "Auto Browse" feature allows users to delegate online tasks to an AI agent directly within a Chrome tab. It is accessed through the Gemini sidebar in the browser, where users provide instructions for the desired action. Once initiated, Gemini will indicate "Task started" and can perform actions such as clicking, inputting text, and scrolling to navigate and complete tasks on websites.

Demonstrations of its capabilities include:

  • Online shopping, such as searching for items on platforms like Etsy, adding them to a cart, applying discount codes, and personalizing items.
  • Scheduling appointments.
  • Completing online forms and gathering tax documents.
  • Obtaining service quotes.
  • Managing bills and subscriptions.
  • Submitting expense reports.
  • Expediting driving license renewals.
  • Booking flights, reserving tickets, and planning vacations.
  • Reordering items.

For certain tasks, such as accessing login information, the AI can, with user permission, interface with Google Password Manager.

Availability and Subscription Tiers

"Auto Browse" is currently available in a preview mode exclusively to subscribers of Google’s AI Pro and AI Ultra plans in the U.S.

  • Google AI Pro subscribers ($20/month) have a daily request cap of 20 tasks.
  • Google AI Ultra subscribers ($250/month) have a daily request cap of 200 tasks.

Multiple tasks can be executed concurrently. Google has not specified a timeline for the feature's wider release to non-paying users or in other countries.

Users can activate "Auto Browse" via an opt-in pop-up or by enabling the 'Let Chrome browse for you' toggle in settings. Once activated, direct prompts in the Gemini sidebar will trigger the automated browser actions.

User Control and Security Measures

Google emphasizes user responsibility and control within the "Auto Browse" environment. A "Take over task" button is available in the browser window, allowing users to intervene and manually control the browser interaction at any point.

Tasks considered more sensitive, such as completing purchases, processing credit card transactions, or posting on social media, are designed to require explicit user oversight and confirmation for the final step. Google includes a disclaimer stating:

"Use Gemini carefully and take control if needed. You are responsible for Gemini’s actions during tasks."

Regarding security, Google acknowledges that "Auto Browse" and similar AI tools are susceptible to prompt injection attacks, which could lead the bot to perform unintended actions if visiting malicious websites. Google states it is developing "layered defenses" to address these threats. General concerns for AI browsers also include the risk of generating incorrect information.

Industry Context and Future Developments

Google's introduction of "Auto Browse" aligns with a broader industry trend of integrating generative AI into web browsing experiences, with competitors like Microsoft, OpenAI, and Perplexity also offering "agentic" AI browser functionalities. Google had previously integrated its Gemini assistant into Chrome, including a dedicated button.

During demonstrations to journalists, the feature's performance showed limitations, with some instances of slow processing of requests or requiring login information for tasks like re-ordering a jacket and searching for discounts. Some initial user reports indicated that Gemini stated it had 'taken over the navigation' but did not perform actual browser clicks, requiring logging out and refreshing the browser. A wider release is pending as Google addresses potential issues.

Google is also developing infrastructure for the feature and Gemini to integrate with third-party platforms. Chrome is set to support the Universal Commerce Protocol (UCP), described as a new open standard for agentic commerce. This protocol, co-developed with companies such as Etsy, Shopify, Target, and Wayfair, is designed to facilitate seamless interaction between AI agents and diverse websites.