Project Mariner: Google’s Revolutionary Web-Browsing AI Agent Expands Access
In a significant move that signals the dawn of a new era in how we interact with the internet, Google announced at its I/O 2025 conference that the Project-Mariner AI agent—the company’s experimental web-browsing AI assistant—is now rolling out to more users and developers. This development represents a fundamental shift in web interaction, with Google’s Project Mariner AI agent poised to transform how we complete online tasks.
What is Project Mariner?
Origins and Development
The Project Mariner AI agent, first unveiled in late 2024, represents Google‘s boldest effort yet to revolutionize web interaction through AI technology. Built on Google’s advanced Gemini 2.0 foundation model, the Project-Mariner AI agent combines sophisticated multimodal understanding and reasoning capabilities to browse and interact with websites on behalf of users.
Key Features and Capabilities
The most striking capability of Google’s Project Mariner AI agent is its ability to perform tasks that typically require human interaction with websites. Users can simply chat with the Mariner AI agent, which then visits websites and takes actions for them—from purchasing tickets to a baseball game to buying groceries online, all without the user ever visiting a third-party website themselves.
According to Google DeepMind’s official page, Project-Mariner possesses three core capabilities:
- Native Multimodality: Understands and reasons across everything on your browser screen, including text, code, images, and forms, while responding to voice instructions and providing visual feedback.
- Browser Interaction: Navigates and interacts with complex websites in real-time, automating repetitive tasks while keeping users in control.
- Advanced Reasoning: Interprets complex instructions, breaks them down into actionable steps, and provides clear visibility into its decision-making process.
Project Mariner’s Latest Updates
Performance Improvements
Perhaps the most significant update announced at Google I/O 2025 is the Project-Mariner AI agent’s ability to multitask. According to a Google spokesperson, the company has updated the Mariner AI agent to run on virtual machines in the cloud, similar to agents from OpenAI and Amazon. This crucial improvement allows users to work on other projects while the Project Mariner AI agent completes tasks in the background—with the ability to handle up to 10 tasks simultaneously.
This represents a major improvement over the previous version, which ran directly in the user’s browser and required them to keep the tab active. As TechCrunch reporter Maxwell Zeff noted in his review, this earlier limitation “defeated the purpose of an AI agent—it would work for you, but you couldn’t do anything else while it was working.”
Subscription Model and Availability
Initially, access to the Project Mariner AI agent will be limited to U.S. subscribers of Google’s new AI Ultra plan, priced at $249.99 per month. Google has indicated that support for more countries is coming soon, along with plans to bring the Project Mariner AI agent’s capabilities to the Gemini API and Vertex AI, enabling developers to build applications powered by the intelligent web-browsing agent.
How Project Mariner Transforms Web Browsing
Real-World Applications
Project Mariner’s potential applications span numerous domains:
- Shopping and E-commerce: Comparing products, reviews, and prices; automating checkout processes; and tracking orders in real-time.
- Research and Information Gathering: Extracting key insights from articles, sourcing relevant papers, and creating comprehensive reports.
- Enterprise Automation: Automating data entry into CRMs, conducting market research by analyzing competitor sites, and streamlining administrative tasks.
- Travel and Event Planning: Finding and booking flights, hotels, and event tickets based on user preferences.
In the coming months, Google plans to integrate the Project-Mariner AI agent with its AI Mode, the company’s AI-powered Google Search experience. Initially, this feature will be limited to Search Labs, Google’s testing ground for search features. Google is already working with partners like Ticketmaster, StubHub, Resy, and Vagaro to power some of these “agentic flows” that will showcase the Mariner AI agent’s capabilities.
Potential Impact on Internet Economy
Web-browsing AI agents like the Project Mariner AI agent have significant implications for the internet economy. As users increasingly delegate tasks to these intelligent AI agents rather than visiting websites themselves, the relationship between vendors and customers could fundamentally change. This shift raises important questions about how websites will adapt their design and functionality to accommodate AI agents like Project-Mariner, as well as potential changes to advertising and revenue models.
Project Mariner vs. Competitor AI Agents
Google isn’t alone in developing web-browsing AI agents. The Project Mariner AI agent competes with several similar tools:
- OpenAI’s Operator: Another experimental web-browsing agent that can perform tasks on websites.
- Amazon’s Nova Act: Amazon’s entry into the web-browsing agent market.
- Anthropic’s Computer Use: A similar tool focused on browsing and using websites.
According to TechCrunch, these tools are all still in experimental stages, with prototypes being “slow and prone to mistakes.” However, Google claims it has taken feedback from early testers to significantly improve the Project-Mariner AI agent’s capabilities.
Future Roadmap for Project Mariner
Integration with Google Search
Beyond the initial rollout to AI Ultra subscribers, Google has ambitious plans for the Project-Mariner AI agent’s future. As mentioned, the company plans to integrate the Mariner AI agent into its AI Mode search experience, potentially transforming how users interact with search results.
Additionally, Google unveiled a demo of another agentic experience called “Agent Mode,” which combines web browsing with research features and integrations, as well as connections with other Google apps. Google has indicated that Ultra subscribers will gain access to Agent Mode on desktop soon.
Development Opportunities
By bringing the Project Mariner AI agent’s capabilities to the Gemini API and Vertex AI, Google is creating opportunities for developers to build applications that leverage these powerful web-browsing capabilities. This could lead to a new ecosystem of AI-powered apps that use the Mariner AI technology to automate various aspects of web interaction.
Some companies—including Browserbase, Cartwheel, and Automation Anywhere—are already experimenting with Project Mariner’s capabilities. Broader developer access is expected to be available “this summer,” according to Google.
Conclusion
At Google I/O 2025, the company finally seems ready to ship the agentic experiences it has been discussing for years. The Project Mariner AI agent, along with Agent Mode and AI Mode, appears poised to fundamentally change how users navigate the web and how vendors interact with their customers online.
While there are still limitations to overcome and ethical considerations to address, Google’s significant updates to the Project Mariner AI agent—particularly its ability to multitask in the background—represent a major step forward in making AI agents practical for everyday use.
As these technologies continue to evolve, the Project Mariner AI agent and similar tools promise to save users time and streamline many aspects of web interaction, potentially marking the beginning of a new paradigm in how we use the internet.
Frequently asked questions.
Answers connected directly to this article and its subject.
01 What is Google's Project Mariner?
The Project Mariner AI agent is Google’s experimental web-browsing assistant that can browse and interact with websites on behalf of users, automating tasks like shopping, research, and booking tickets without users having to visit third-party websites themselves.
02 How much does it cost to access Project Mariner?
Currently, the Project Mariner AI agent is available to U.S. subscribers of Google’s AI Ultra plan, which costs $249.99 per month. Google has announced plans to expand availability of the Mariner AI agent to more countries soon.
03 How is Project Mariner different from regular web browsing?
The Project Mariner AI agent allows users to delegate tasks to an intelligent assistant instead of visiting websites and completing those tasks themselves. Users simply chat with the Mariner AI agent, which then navigates websites and takes actions on their behalf.
04 What improvements has Google made to Project Mariner?
The latest version of the Project Mariner AI agent runs on virtual machines in the cloud, allowing users to work on other projects while it completes tasks in the background. Google’s Mariner AI agent can now handle up to 10 tasks simultaneously, a significant improvement over the previous version.
05 How does Project Mariner compare to competitors like OpenAI's Operator?
Project Mariner competes with several similar tools, including OpenAI’s Operator, Amazon’s Nova Act, and Anthropic’s Computer Use. While all these tools are still experimental and have limitations, Google claims to have made significant improvements to Project Mariner based on early user feedback.
