Google’s Project Mariner is an experimental AI agent developed by DeepMind, designed to autonomously navigate and interact with web pages on behalf of users. Operating as a Chrome browser extension, Mariner can read, click, scroll, and input information into forms, effectively automating various online tasks.

Built upon the Gemini 2.0 AI model, Mariner possesses advanced multimodal understanding, enabling it to process diverse web elements such as text, images, code, and forms. This comprehensive comprehension allows Mariner to execute complex tasks, like adding items to an online shopping cart or planning travel itineraries, while providing users with real-time updates and maintaining control over the process.

Currently, Project Mariner is in a testing phase with a select group of users. Google is focusing on ensuring user safety by implementing measures to prevent unauthorized actions, such as making purchases or accepting cookies without explicit user consent. Additionally, Mariner is designed to prioritize user instructions over potentially malicious prompts from external sources, enhancing security against fraud and phishing attempts.

By integrating advanced AI capabilities into the browsing experience, Project Mariner represents a significant step toward more intuitive and efficient human-computer interactions, aiming to simplify and streamline online activities for users.

For a visual overview of Project Mariner, you can watch the following video: