AI RankiaAI agent readiness: can an agent use your site?
One scan of your domain answers with evidence: what GPTBot gets, which agent files exist, which WebMCP tools the page registers after render and, for stores, the agentic checkout signals. Each MISSING check comes with a proposed fix.
The full scan runs inside AI Rankia on credits, from Monitor ($99/mo). It checks a website, not whether your company is ready to deploy AI agents.
- PASSai bot accessYour homepage answers 200 OK to the GPTBot crawler.
- PARTIALai rules in robots.txtrobots.txt names no AI bots by name.
- MISSINGllms.txtReturns 200 OK but serves your normal web page (HTML).
- MISSINGUCP manifestUCP manifest returned HTTP 404.
- PASSACP prerequisites (OpenAI)Prerequisites met: Shopify checkout + product data.
- MISSINGWebMCP toolsNo agent-executable tools.
Crawl, render, score by profile
Crawl like an agent
robots.txt, sitemap, llms.txt, /.well-known files and your homepage as GPTBot. Firewall blocks are retried via a mobile proxy.
Render in headless Chrome
Schema after JavaScript, language, one H1, and the WebMCP tools the page registers with document.modelContext or as forms.
Score by profile
Stores weigh commerce; other sites weigh discovery and machine-readability. 0 to 100.
Proof for every check, and the fix
See what we found, the proof and the fix
Each check is PASS, PARTIAL or MISSING, with what is on your site now, how we tested it and the content the URL returned. One click opens the real file to verify it yourself.
- Soft 404s caught: a 200 that serves your homepage is a miss
- One-click robots.txt fix through Cloudflare
- Scan history, and any scan as a .pptx deck
robots.txt exists (4 lines) but names no AI bots by name (GPTBot / ClaudeBot / PerplexityBot…).
On your site now · See the real file ↗
GET /robots.txt → parse User-agent blocks
proof — what the file actually contains
User-agent: * Disallow: /cart Disallow: /checkout Sitemap: https://example-store.com/sitemap.xml
Proposed · not on your site yet
User-agent: GPTBot Allow: / User-agent: ClaudeBot Allow: /
Give agents tools to call, not pages to scrape
We host a small script that registers WebMCP tools for your site with document.modelContext.registerTool(), shaped by the scan: catalog and cart for stores, search and a next step for other sites.
- Install with Google Tag Manager in one click (25 credits) or paste one line
- Product and page data come only from what is already on your pages
- Card payments stay on your own checkout
Store
- read
search_catalogName, price, availability and URL of matching products. - read
get_productFull details from the structured data on the page. - write
add_to_cartRuns on the guest session, no login. · if a cart was found - read
view_cartContents and total of the cart. · if a cart was found - nav
start_checkoutSends the shopper to your own checkout. · if a cart was found - write
pay_cryptox402 payment details, USDC to your wallet. · only with a wallet saved
Brand or SaaS site
- read
search_contentRelevant sections and links of your pages. - read
get_pageHeadings, key text and links as clean text. - nav
primary_actionYour demo, trial or contact link.
Agentic checkout and agent payments
Content & GEO: can AI read one page?
What the scan can't see
- It reads your homepage, not your whole site.
Plus the standard files and, on stores, one product page sampled from the sitemap. Deeper pages are not rendered.
- ACP enrollment is invisible from outside.
We check the prerequisites: a payment rail and product data. Whether OpenAI has switched you on, no scan can tell.
- Our render browser has no native WebMCP.
A listener we inject before your scripts records each
document.modelContext.registerTool()call, and we count<form toolname tooldescription>after render. Tools added later, or only after a click, can be missed.A listener we inject records eachregisterTool()call ondocument.modelContext; tools added after a click can be missed. - A score is not a ranking factor.
Google: you “don't need to create new machine readable files, AI text files, markup, or Markdown” to appear in Search (guide).
A scan costs credits. Fixes cost more.
Source: AI Rankia credit costs, Oct 13, 2026. Reopening a past scan or its .pptx costs nothing.
Four ways to check agent readiness
Many free checkers exist; the two most cited come from Cloudflare and Chrome. Cloudflare's scanner checks more agent standards than we do, commerce included. Lighthouse checks llms.txt, WebMCP and agent accessibility. Compared with those two, we add the homepage as GPTBot sees it, schema after JavaScript, a score weighted for stores, and a robots.txt fix we deploy.
| Feature | AI Rankia scan | Cloudflare isitagentready.com | Lighthouse Agentic Browsing (Chrome) | By hand (curl + DevTools) |
|---|---|---|---|---|
| robots.txt AI bot rules | Yes | Yes | Not found advertised | curl |
| llms.txt | Yes | Not found advertised | Yes | curl |
| Homepage fetched as GPTBot | Yes | Not found advertised | Not found advertised | curl |
| Schema after JS | Yes | Not found advertised | Not found advertised | DevTools |
| WebMCP tools | document.modelContext + forms, via our listener | Yes | Chrome 150+, origin trial | DevTools console |
| Agent accessibility, layout shift | Not found advertised | Not found advertised | Yes | DevTools |
| MCP card, Agent Skills, OAuth | Agent skill file, unscored | Yes | Not found advertised | curl each file |
| UCP, A2A, x402 | Stores | Also MPP, ACP | Not found advertised | curl each file |
| Fix | robots.txt via Cloudflare; WebMCP via GTM | Instructions for your coding agent | Not found advertised | Not found advertised |
| Price | 5 credits per scan | Free | Free | Free |
Yes advertisedNot found advertised not found advertised
Checked on on each vendor's public pages and docs. ✗ means we did not find it advertised on that date, not that it does not exist.
Lighthouse: “The Agentic Browsing category and WebMCP support are experimental and based on proposed standards.” Cloudflare labels its improvement tips “AI-generated recommendations”. Our render browser has no native WebMCP: we capture document.modelContext.registerTool() calls with our own listener; see Limits.
Sources and notes
Lighthouse: “The Agentic Browsing category and WebMCP support are experimental and based on proposed standards.” Cloudflare labels its improvement tips “AI-generated recommendations”. Our render browser has no native WebMCP: we capture document.modelContext.registerTool() calls with our own listener; see Limits.
Questions people ask
What is AI agent readiness?
What is WebMCP?
Can ChatGPT buy from my store?
What is agentic SEO?
Is there a free agent readiness check?
Agent-ready is half the job. Do AI engines name you?
The free scan asks AI engines your buyers’ questions: do they name you, who instead, and which pages they cite. No signup.
