Google Search Advocate John Mueller recently confirmed that Google currently has no implementation for `llms.txt`, dismissing the proposed standard as "speculative" for the time being. Instead, Mueller expressed a strong preference for WebMCP, a web-focused implementation of the Model Context Protocol (MCP) backed by Google and Anthropic.
Technical Context
The `llms.txt` proposal recently gained traction as an AI-era equivalent to `robots.txt`. It is designed to provide large language models with a standardized, markdown-formatted summary of a website's content and structure, relying on passive, static file parsing during the crawling phase. Conversely, WebMCP (built on the Model Context Protocol) is a dynamic, API-driven approach. It allows AI agents to actively query and retrieve context, tools, and prompts from a server in real-time. By favoring WebMCP, Google is signaling a shift away from passive text files toward active, client-server architectures for AI data ingestion.
Why It Matters
For engineering and SEO teams, this is a clear signal to adjust resource allocation. Implementing `llms.txt` might still hold value for smaller, independent scrapers or niche AI projects, but it will not currently influence how Google's AI Overviews or Gemini agents index and understand your site. Developers should avoid treating `llms.txt` as a silver bullet for "AI SEO." Instead, focus should shift toward building MCP servers. Exposing site data, APIs, and structured context through WebMCP ensures compatibility with the architecture Google is actively endorsing.
What to Watch Next
Monitor Google's official developer documentation for formal WebMCP integration guidelines, particularly how it might intersect with Googlebot crawling or Gemini's agentic workflows. Additionally, watch to see if other major players like OpenAI adopt MCP or stick to proprietary crawling heuristics, which will determine whether WebMCP becomes the undisputed industry standard for web-to-LLM interfacing.
