Agent Readiness Checker
Is your site agent-ready? See what AI agents can reach and read today, and which of the new agent standards you publish.
| Check | Result | What we found |
|---|
| Standard | Result | What we found |
|---|
What is agent readiness?
An AI agent is software that visits websites for a person. It fetches a page, reads it, and sometimes takes the next step: compares prices, fills a form, calls an API. Agent readiness is how well your site works for that kind of visitor.
It comes down to three questions. May agents reach the site at all? Can they read what they fetch? And does the site publish the files that tell an agent what it can do there?
The first two are old problems with a new audience, and they are where almost all of the value is today. The third is a set of young standards, most of them proposals, that few sites publish and few agents rely on yet.
What this agent readiness checker checks
The result comes in two groups, and they are kept apart on purpose.
Read by agents today. Crawler rules, content without JavaScript, page basics, the sitemap, structured data, Markdown for agents and llms.txt. These have a visible effect now. Problems here are worth fixing.
Emerging standards. WebMCP, MCP server cards, A2A agent cards, Agent Skills, Content Signals and the rest. The checker tells you which ones your site publishes. A missing one is not a fault.
There is no score. A single number would let a missing payment protocol pull down a site that agents read perfectly well, and that tells you nothing useful.
What AI agents read today
AI crawler rules
Every well-behaved agent starts with robots.txt. If the file blocks it, nothing else on this page matters. The checker tests the page against each search and AI crawler and names the ones that are shut out. To see the rule behind a block, open the site in the Robots.txt Tester.
Content without JavaScript
Most AI crawlers do not run scripts. They read the HTML the server sends and stop there. If your content is assembled in the browser, they get an empty shell. The checker counts the words in the server HTML, which is the simplest honest measure of what such a crawler sees.
Page basics
A title, one main heading and a language. An agent uses them to decide what the page is before it reads the rest.
Sitemap
A sitemap lists the pages you want found. Without one, a crawler has to discover them by following links, and it will miss some.
Structured data
JSON-LD states the facts of a page in machine terms: who the organization is, what the product costs, what the questions and answers are. The Schema Validator shows what yours says and what is missing.
Markdown for agents
An agent can ask for a page as Markdown by sending Accept: text/markdown. A server that supports this returns the content without the navigation, scripts and layout around it, which makes the page much lighter to read. It is optional, and it is one of the few new ideas with a measurable effect.
llms.txt
A short file that points AI models at the pages that explain your site. It is used today mostly by coding tools. The llms.txt Checker validates it.
Emerging agent standards, explained
These are the standards the checker looks for in its second group. Each solves a real problem for some kind of site. None of them is needed by a site that has nothing for an agent to call.
| Standard | What it is | Where it lives | Who it is for |
|---|---|---|---|
| WebMCP | A way for a page to offer tools to an agent working in the browser | navigator.modelContext in the page | Sites with actions an agent could take: search, booking, checkout |
| MCP server card | A pointer to the site's MCP server | /.well-known/mcp/server-card.json | Products that expose tools or data to AI assistants |
| A2A agent card | A description of an AI agent the site runs | /.well-known/agent-card.json | Companies that operate their own agent |
| Agent Skills | An index of instructions an agent can load to use the site well | /.well-known/agent-skills/index.json | Products with workflows worth teaching an agent |
| Content Signals | A line that says how your content may be used: search, AI answers, AI training | Content-Signal in robots.txt | Any site that wants to state its terms |
| Link headers | Pointers to machine-readable resources, sent with the page | The Link response header | Sites with an API catalog or other files to advertise |
| API catalog | A list of the site's APIs and their specifications | /.well-known/api-catalog | Sites with a public API |
| OAuth discovery | Files that tell an agent how to sign in on a user's behalf | /.well-known/oauth-authorization-server and related | Sites with accounts and an API |
| Web Bot Auth | Public keys a bot uses to sign its requests, so sites can verify it | /.well-known/http-message-signatures-directory | Companies that run their own crawler or agent |
| UCP and ACP | Two protocols that let an agent complete a purchase | /.well-known/ucp, /.well-known/acp.json | Online shops |
What is WebMCP?
WebMCP is a proposed browser standard. It lets a page hand an agent a list of tools, each with a name, a description and the inputs it takes. The agent then calls a tool directly, which is faster and more reliable than working out which button to press. The checker sees WebMCP only when the registration code is in the page HTML itself.
MCP server card and A2A agent card
Both are small JSON files that help agents find things. The MCP server card says the site has an MCP server and where to connect. The A2A agent card describes an agent the site runs, so that other agents can work with it. Without the card, someone has to enter the address by hand.
Content Signals
One line in robots.txt, for example Content-Signal: search=yes, ai-input=yes, ai-train=no. It separates being found in search, being used in an AI answer, and being used to train a model. It states a preference. It does not enforce one.
The payment protocols
UCP and ACP are checked because each has a file at a fixed address. Two others, x402 and MPP, work through the payment flow itself and cannot be seen from the outside, so this checker does not report on them.
Does agent readiness matter for AI visibility?
The first group does, plainly. An assistant cannot quote a page it cannot fetch or read.
The second group is different. There is no public evidence that publishing an MCP server card, an agent card or a WebMCP tool makes an assistant cite or recommend a site more often. These standards are about letting an agent do something on your site. They are not signals about whether your site is worth mentioning.
So the order is simple. Make sure agents can reach and read you. Then, if your product has something an agent could use, look at the standards that fit it. A site that sells nothing through an API loses nothing by leaving the payment protocols alone.
This is also where agentic SEO stands today: mostly ordinary technical SEO, done with a second reader in mind.
How to make your site agent-ready
- Open the door. Check that
robots.txtdoes not block the AI crawlers you want, and that your firewall does not block them either. The AI Crawlability Checker tests the firewall. - Put the content in the HTML. If the page is empty until a script runs, render it on the server.
- Publish a sitemap and name it in
robots.txt. - Add structured data. Start with an Organization item on the homepage.
- Consider Markdown and llms.txt. Both are cheap and both are optional.
- Add the newer standards only where they fit. An MCP server card if you run an MCP server. WebMCP if agents could act on your pages. Commerce protocols if you sell online.
Frequently asked questions
What is agent readiness?
Agent readiness is how well a website works for AI agents: software that fetches pages, reads them and sometimes acts on them for a person. It covers whether agents may reach the site, whether they can read it, and whether the site publishes the files agents look for.
How do I check if my site is agent-ready?
Enter your site in the checker above. It tests what agents read today, such as crawler rules, content without JavaScript and structured data, then lists which of the newer agent standards the site publishes.
Why is there no score?
Because a single number would mix things that matter with things that do not yet. A site that agents can reach and read is in good shape. Publishing an MCP server card on top of that does not make it twice as ready. The checker shows the two groups apart and lets you judge.
What is WebMCP?
WebMCP is a proposed browser standard that lets a web page offer tools to an AI agent working in the browser. The page registers functions, such as a search or a booking step, through navigator.modelContext, and the agent calls them directly without clicking through the interface.
What is an MCP server card?
A small JSON file at /.well-known/mcp/server-card.json that tells an agent the site runs an MCP server and where it is. MCP, the Model Context Protocol, is the common way for AI assistants to connect to tools and data. The card saves the agent from being configured by hand.
What is an A2A agent card?
A JSON file at /.well-known/agent-card.json that describes an AI agent the site operates: what it can do and how to talk to it. It comes from the Agent2Agent protocol, which lets agents from different makers work together.
Does agent readiness help me show up in ChatGPT?
Not by itself. Being reachable and readable is the condition for being used at all. The newer standards have no shown effect on whether an assistant cites or recommends you. What moves that is content worth citing and mentions on sites the assistants already trust.
Does a small site need MCP, WebMCP or agent cards?
Usually not. Those standards serve sites that offer an API, a tool or a checkout an agent can use. A content or marketing site gets nearly all the benefit from the first group: open crawler rules, readable HTML, a sitemap and structured data.
What is agentic SEO?
Agentic SEO is the work of making a site usable by AI agents as well as by search engines. In practice most of it is ordinary technical SEO, with a few additions such as Markdown for agents and llms.txt.
Why does my result differ from other agent readiness checkers?
Most checkers score every standard equally, so a site without commerce protocols or OAuth discovery loses points even if it has no use for them. This checker reports those as information only. The facts found should match; the judgement is what differs.
An agent-ready site is one that agents can reach and read. Whether AI assistants name your brand when someone asks is a separate question, and the AI Visibility check answers it.