AI News Bureau
Written by: Abid Hussain Barlaskar
Updated 7:25 PM UTC, Fri May 30, 2025
Google is accelerating its push into agentic AI with the wider rollout of Project Mariner, its experimental AI web-browsing assistant, the company announced at its I/O 2025 conference.
Initially launched in late 2024, Project Mariner allows users to delegate online tasks — such as booking tickets or shopping — to an AI agent, eliminating the need for third-party websites. Google says the latest version of the agent can now handle up to 10 tasks at once, thanks to a major infrastructure shift that runs it on cloud-based virtual machines.
U.S. subscribers to Google’s new $249.99/month AI Ultra plan will be the first to access the upgraded Mariner, with international support coming soon. Developers will also gain access via the Gemini API and Vertex AI, enabling deeper integrations into apps and services.
Project Mariner competes with similar offerings from OpenAI, Amazon, and Anthropic, but Google claims improved performance and multitasking based on early user feedback. Previously limited by its dependency on a user’s browser, Mariner now runs in the background, freeing up users to work elsewhere.
Google also unveiled a broader agentic ecosystem, including “Agent Mode,” which combines web browsing with integrated Google app support, and “AI Mode” for Search — both aimed at redefining how users interact with the web.
Mariner’s rollout comes with partnerships from platforms such as Ticketmaster, Resy and StubHub to enable agent-driven commerce experiences.