Google Launches Project Mariner, Its AI Web-Browsing Agent, to Select Users

Google has officially announced the rollout of Project Mariner, its experimental AI agent designed to browse and interact with websites, making it available to a wider audience of users and developers. This announcement was made during Google I/O 2025. The company has also significantly updated Project Mariner’s capabilities, enabling it to handle nearly a dozen tasks simultaneously.

Initially, Project Mariner will be accessible to U.S. subscribers of Google’s new $249.99-per-month AI Ultra plan. Google plans to expand support to additional countries soon. Furthermore, Project Mariner’s functionalities are being integrated into the Gemini API and Vertex AI, empowering developers to create applications driven by this AI agent.

First introduced in late 2024, Project Mariner aims to transform how users engage with the internet through AI agents. Google Search leaders envision a future where users delegate tasks to AI agents rather than directly interacting with websites themselves.

For example, Project Mariner allows users to purchase baseball tickets or order groceries online simply by interacting with Google’s AI agent, which then navigates websites and completes the transactions on their behalf.

Project Mariner enters a competitive landscape alongside other web-browsing AI agents such as OpenAI’s Operator, Amazon’s Nova Act, and Anthropic’s Computer Use. Currently, these tools are in experimental phases and have demonstrated some limitations in speed and accuracy.

Google has incorporated feedback from early testers to refine Project Mariner’s performance. A Google spokesperson confirmed that the agent now operates on virtual machines in the cloud, similar to agents from OpenAI and Amazon. This improvement allows users to continue working on other tasks while Project Mariner operates in the background, handling up to 10 tasks concurrently.

This enhancement marks a significant improvement over the initial version of Project Mariner, which required users to keep the agent’s browser tab active, preventing them from using other applications. This limitation undermined the fundamental purpose of an AI agent designed to enhance user productivity.

In the near future, Google intends to integrate Project Mariner into AI Mode, its AI-driven Google Search experience. This feature will initially be available through Search Labs, Google’s platform for testing new search functionalities. Google is partnering with Ticketmaster, StubHub, Resy, and Vagaro to facilitate these agent-driven interactions.

Google also previewed “Agent Mode,” an upcoming feature that combines web browsing with research capabilities and integrations with other Google applications. Agent Mode will soon be available to Ultra subscribers on desktop.

Google’s recent announcements at I/O indicate a shift towards delivering tangible agent-based experiences. Project Mariner, Agent Mode, and AI Mode have the potential to reshape how users interact with the web and how businesses engage with their customers. The broader implications of web-browsing agents for the internet economy are substantial, and Google appears ready to deploy these technologies on a large scale.