Free Tool

Can an AI agent find you,read you, and buy from you?

A deterministic scorecard for the seven surfaces agents actually touch — llms.txt, Content Signals, sitemap, product data, agentic checkout, .well-known discovery, markdown mirrors — scored over the applicable denominator: a check that doesn't apply to what your site is gets excluded, not failed. For what you declared yourself to be, here is the honest number.

Step 1 — declare what your site is

The declaration scopes the denominator: a not-applicable check is excluded, not failed. Declaring wrong to inflate the score is itself detected and gated.

Step 2 — paste your files

Nothing you paste leaves your browser. The scoring runs client-side on exactly what you give it, so the same inputs produce the same number every time. Leave a box blank if the file genuinely does not exist — that is a real signal, not a gap.

.well-known files — optional

Leave every box blank if you publish none; that is the common case and it is a real answer, not a gap. Paste the file body, not the URL.

Why the denominator is the whole point

Composite agent-readiness scores run API, OAuth, and commerce checks against every site — which quietly makes ~50/100 the ceiling for a content business and teaches owners to chase points that don't apply to them. This scorecard inverts that: declare what your site is, and only the checks that apply count. The one thing you can't do is declare yourself wrong to make the number bigger — that's detected from your own artifacts and reads MISDECLARED until the declaration matches the evidence.

FAQ

What does the Agent-Readiness Scorecard check?

Seven agent-facing surfaces: llms.txt, robots.txt Content Signals, your sitemap, JSON-LD Product data, ACP/UCP agentic checkout readiness, .well-known discovery (MCP server card, ai-plugin.json), and markdown mirrors / content negotiation. Each is marked 0–2 from the actual file contents, weighted by what your site is.

Why doesn't a content site get penalized for missing checkout checks?

Because a not-applicable check is excluded from the denominator, not failed. Composite scorecards that run commerce checks against a blog cap it near 50 forever — that's a misleading number, not a low one. Here the verdict is: for what you declared yourself to be, this is the honest score.

What happens if I declare my site as 'content' to hide a weak storefront?

The scorecard reads MISDECLARED — regardless of the score. Product JSON-LD with offers, or commerce URLs in your sitemap corroborated by payment-stack markers, prove the site sells; excluding the commerce checks then inflates the number, so the gate floors the verdict until the declaration matches the evidence. It fires only in the inflating direction.

Does the Agent-Readiness Scorecard send my site anywhere?

No. You paste robots.txt, llms.txt, your sitemap, the homepage source, one product page, and any .well-known files you publish, and the scoring runs client-side in your own browser — nothing is uploaded, stored, or logged. That is also why the number is reproducible: the same inputs produce the same score every time, on our machine or yours.

I scored above 75 but it still says HALF-LIT. Why?

The dark-surface gate. LIT TO AGENTS is a claim about your whole agent-facing surface, so a top-band score is held at HALF-LIT while any applicable check reads 0/2 — one entirely absent surface is a wall an agent hits no matter how strong the rest is. It names which surface in the verdict panel. Like every gate here it is worsen-only: it can lower a verdict, never raise one.

Is the score comparable to Cloudflare's isitagentready.com?

No, deliberately. Cloudflare's composite runs every category against every site, which makes ~50/100 the de facto ceiling for content sites. This scorecard scopes the denominator to your declared site type, so the number means the same thing for a blog and a storefront: how much of what applies to you is actually in place.

What should I fix first if I score badly?

The scorecard names it: the applicable check losing the most weighted points. For most storefronts that's structured product data or an agent-readable checkout surface; for most content sites it's markdown mirrors. One fix first, then re-run. If it flags structured product data, check you pasted a product page and not just the homepage — most sites keep their Product schema on product pages.

Scored LIT and sell online? Go deeper.

This scorecard grades the site's agent-facing surface. The item-by-item feed audit — whether an AI shopping agent can machine-read, trust, and act on each product — is the Agentic-Commerce Product-Feed Readiness Gate ($79). For how AI engines describe you, start with the AEO Citation Audit Kit ($79) and fix it with the GEO / AI Visibility Playbook ($149).

Sold by RedHub AI LLC · Secured by Stripe · redhub.ai