What is agent readiness? Definition and checklist

Agent readiness is how well a website can be found, read and used by AI agents acting for people, from assistants that research vendors to browsers that fill in forms.

How it works

There is no single standard for agent readiness. As of October 2026, two public scoring tools show what people mean by it.

Cloudflare's Agent Readiness score, launched in April 2026, checks four areas (Cloudflare):

• Discoverability: robots.txt, sitemap.xml and Link headers.
• Content: whether the site can serve Markdown to agents. llms.txt is an optional extra check, not part of the default scan.
• Bot access control: Content Signals, AI bot rules in robots.txt and Web Bot Auth.
• Capabilities: machine-readable entry points such as an API catalog, OAuth discovery, an MCP server card and WebMCP. Agentic commerce protocols are detected but not scored.

Chrome's Lighthouse has an experimental Agentic Browsing category that needs Chrome 150 or later. Instead of a 0–100 score it shows how many checks a page passes, covering registered WebMCP tools, llms.txt, accessibility for agents and layout stability (Lighthouse). Lighthouse notes that agents "rely on the accessibility tree as their primary data model", so unlabeled buttons and inputs hurt agents as much as screen reader users.

When Cloudflare Radar scanned popular domains, 78% had a robots.txt file, but most were written for search crawlers; only 4% declared AI usage preferences, and Markdown negotiation passed on 3.9% of sites.

Why it matters for your website

Buyers increasingly ask an assistant to shortlist vendors, compare prices or book a demo. An agent that cannot read your pricing, or gets stuck on an unlabeled form, moves on to a site it can use.

Start with the basics. Google says pages need no special files or markup to appear in its AI features; they must be indexed and eligible for a snippet (Google Search Central). So the first layer is plain, crawlable text. The newer layers, such as llms.txt, Markdown versions of pages and WebMCP tools, are drafts or conventions: worth trying, not worth rebuilding a site for.

A practical way to think about it:

1. Readable: key facts in HTML text, not images or PDFs.
2. Understandable: clear headings, prices, limits and structured data.
3. Usable: labeled forms and stable layouts, then experimental tools for key actions.

Example

A software company runs a quick audit. Its robots.txt blocks every bot by default, the pricing table is an image, and the demo form's fields have no labels. It allows the assistants it wants in robots.txt, rebuilds the pricing table as HTML, labels every field, and publishes a short llms.txt. Its next scan passes more checks, and an assistant asked about the vendor can now quote the real prices.

The free agent readiness checker and the agent-ready website checklist walk through the same steps.

Related terms

• AI agent: software that plans and takes steps toward a goal.
• Agentic browser: a browser that can act on pages for its user.
• WebMCP: a draft standard for exposing site actions as tools.
• llms.txt: a proposed index of a site's key pages for AI tools.
• robots.txt: the file that tells crawlers where they may go.
• Structured data: machine-readable facts inside your pages.