# What AgentReady is, and what it actually tells you

Someone asks ChatGPT which accounting software to use. Or asks Claude to compare three suppliers. Or tells Perplexity to summarise what a company does before a meeting. In every one of those, a machine reads your website and decides what it says about you.

It does not read it the way a person does. There is no scroll, no design, no patience for a cookie banner. It fetches HTML, throws away what it cannot use, and keeps whatever is left. If what is left is thin, the answer about you is thin.

AgentReady measures what that machine finds.

## What it does

Paste a URL. In a few seconds you get a score out of 100, a converted Markdown version of the page, and a list of what to change, ordered by what matters most. No signup, no card.

The score comes from 23 individual checks across five dimensions:

- **AI discoverability** — llms.txt, robots.txt, whether the AI crawlers are actually allowed in, sitemap, Markdown negotiation
- **Content efficiency** — how much of your page is content and how much is scaffolding, measured in tokens
- **Semantic HTML** — headings that make sense, real `<article>` and `<main>`, alt text, how deep the div nesting goes
- **Structured data** — Schema.org, Open Graph, meta description, canonical, declared language
- **Accessibility for machines** — whether the page works without JavaScript, its weight, where the content sits in the source

Not all 23 count the same. Each is weighted by how good the evidence is that it changes what an agent sees: something documented by the model providers weighs more than something the industry currently believes but has not demonstrated. The rubric is versioned, so a score always says which ruler produced it.

## What you get back

**A score and a grade**, with the breakdown. Not a number on its own — you see which checks passed, which failed, and what each one is worth.

**Your page as Markdown**, the way an agent would keep it after discarding the rest. Reading it is usually the moment it clicks: everything you thought was on the page and is not.

**Recommendations you can act on**, specific to your CMS where we can detect it, ordered by impact rather than by how easy they are to write.

**A badge** you can embed, signed and independently verifiable. Anyone can check the signature without trusting us.

There are also 14 free tools for the individual pieces: generators for llms.txt, robots.txt with AI directives, Content-Signal, meta tags and the well-known descriptors for MCP, A2A and OAuth discovery; and checkers for what you already have.

## We measure ourselves too

agentready.md scores 94 out of 100 on its own rubric, and the number on the home page is read from the database, not typed in. When we lose points, the site says so.

That has cost us. Removing Cloudflare's managed robots.txt dropped our score from 93 to 90, because it turned out the edge had been serving a file we had not written while our own origin returned 404. The honest number was the lower one.

A tool that scores websites and cannot pass its own test is worth exactly nothing.

## Where to start

Run your homepage through it. Then run the page you actually care about — the pricing page, the one describing what you sell — because that is the one an agent will be asked about.

If the Markdown that comes back is shorter than you expected, that is the finding. Everything else is detail.

[Analyse a URL](/) · [Browse the tools](/tools) · [See the leaderboard](/leaderboard)