Google Takes a Step Toward an Internet Built for AI Agents

AI agents are moving beyond answering questions — they’re starting to complete tasks like booking flights or making purchases. This week, Google unveiled WebMCP, a protocol designed to help websites keep up with this shift.

A Google spokesperson told ADWEEK the protocol is necessary because it provides the connection that AI agents need to perform more complex tasks more accurately on the open web.

The protocol is currently available for early preview, which means developers can test the features before it officially launches.

“The goal is to create a standard way for exposing structured tools, ensuring AI agents can perform actions on your side with increased speed, reliability, and precision,” wrote André Cipriani Bandarra, staff developer relations engineer at Google, in a blog post.

Most AI agents currently navigate websites the way a human would, by clicking through pages, scanning buttons, and attempting to complete checkout or make a booking step-by-step. This clunky process is prone to errors.

According to Google’s blog post, WebMCP lets websites expose structured actions for AI agents — essentially the list of predefined tasks they can perform — giving agents a more reliable way to interact with a site instead of guessing which buttons to click or forms to fill.

“By defining these tools, you tell agents how and where to interact with your site. This direct communication channel eliminates ambiguity and allows for faster, more robust agent workflows,” Bandarra wrote.

The protocol could help ecommerce sites guide agents through product searches and checkout flows more reliably, enable travel providers to support flight searches and bookings, and allow customer support systems to accept more detailed, automatically completed tickets, per the blog.

Tech companies are building the infrastructure for agentic AI

Google’s announcement comes amid similar efforts across the industry.

Anthropic has promoted its Model Context Protocol (MCP) for connecting AI systems to external tools and data. Amazon has introduced its own framework for agent interactions across its adtech stack. Trade groups and adtech companies are also rolling out standards, including AdCP, aimed at structuring how AI handles campaign data and media workflows.

WebMCP is designed to play nicely with other standards. The Google spokesperson described it as a complementary, open protocol.

While Anthropic’s MCP requires backend server-side integration, the spokesperson said, WebMCP lets websites use their existing stack for agentic capability.

Many organizations will likely support both to ensure their high-value functionality is accessible wherever users engage with AI agents, the spokesperson added.

All of these moves underscore how quickly the web is being retooled for an agent-driven future.

But marketers might also find new challenges getting these competing protocols to interoperate. If not, all of these protocols will simply add more fragmentation. Still, if AI agents are going to transact, search, and execute on behalf of people, web infrastructure needs to evolve.