As generative AI moves into a critical phase of "automated execution," tech giant Google is accelerating the integration of its internal technology stack. Recent reports indicate that Google has announced the shutdown of its experimental web automation project, Project Mariner, on May 4th. Although this independent project has come to an end, its core technologies have not been shelved but have instead been fully integrated into key product lines such as Gemini Agent and AI Mode.
As an important attempt by Google in the field of AI agents, Project Mariner, which was introduced at the end of 2024, has centered around the selling point of "replacing user operations." It can understand complex cross-website instructions and autonomously execute multi-step tasks. In recent technical iterations, the project has demonstrated the powerful capability to efficiently handle up to 10 concurrent tasks, setting a new technical benchmark for web proxy tools.
This shutdown seems more like a "transition." Over the past year, Google has been working to integrate Mariner's automation features into its own ecosystem. Following this, the Chrome browser showcased a new AI feature called "auto-browse," which also focuses on the automated execution of multi-step tasks.
Industry analysts believe that these actions are a direct response by Google to competing web proxy tools in the market. By transforming scattered experimental projects into built-in capabilities of core products, Google is attempting to build a stronger moat in the competition for AI agents, leveraging its large user base and browser ecosystem.
