Google launches Gemini Intelligence to enable AI agents to perform tasks directly on websites
Google has introduced Gemini Intelligence to enable AI agents to perform automated tasks directly on websites using new standardized protocols.
1. The Shift to Agentic AI
Google unveiled Gemini Intelligence on May 12, 2026, introducing an AI-driven layer integrated into the Android operating system across various devices, including the new Googlebook laptop. This technology represents a transition toward "agentic" systems, where AI agents perform tasks on behalf of users—such as scheduling meetings or visualizing furniture—without the user needing to manually navigate websites. This shift challenges traditional search engine optimization (SEO) models, as AI agents increasingly visit and interact with sites directly rather than directing human traffic to them.
2. Technical Architecture and Protocols
The functionality of Gemini Intelligence relies on two primary protocols designed to facilitate seamless interaction between AI agents and websites. WebMCP, co-developed by Google and Microsoft, allows websites to declare functions as structured tools, enabling agents to execute specific tasks like inventory searches or checkout processes. Complementing this is the Universal Commerce Protocol (UCP), launched in January and supported by companies including Shopify, Walmart, Target, and PayPal. UCP provides a standardized language for agents to manage product discovery, cart building, and order completion. These protocols aim to increase efficiency by allowing agents to bypass the need to parse full HTML documents.
3. Preparing for an Agentic Web
As the operating system evolves into the primary search layer, the role of websites is shifting from consumer destinations to functional backends. To adapt, businesses are encouraged to audit their digital assets to ensure that critical actions, such as booking flows or checkout processes, are accessible to AI agents. Experts suggest monitoring "Agentic Browsing" scores to evaluate how effectively an AI can utilize a site. While the method of interaction is changing, maintaining visibility remains essential, as agents must still be able to find and trust a brand before they can perform tasks on its behalf.