Methodology v1.0

For AI agents

People ask us whether their site is visible to ChatGPT, Perplexity, Claude and Gemini. More and more, an AI agent asks on their behalf. The same scan that powers this site is available to agents and scripts: free, no signup.

What an agent can answer with it

  • "Is GPTBot blocked on example.com?"
  • "Is this site ready for AI search engines?"
  • "What should I fix so Perplexity can read my site?"
  • "Does this domain have llms.txt, a sitemap and structured data?"

Quick start

One HTTP GET returns JSON. No API key. CORS is open.

curl "https://aiocompatible-com.vercel.app/api/v1/analyze?url=example.com"

Optional parameter: category = generic, blog, ecommerce, saas or corporate (changes which schema types count as required). Key response fields:

FieldMeaning
overallScore0-100 score. 85+ Gold, 75-84 Silver, 65-74 Bronze for the badge.
subScorescrawlability, parseability, citationReadiness (each 0-100)
botsStatusArray of {bot, status}: Allowed or Blocked in robots.txt for GPTBot, PerplexityBot, ClaudeBot, Google-Extended, Applebot-Extended
llmsTxtExists, sitemapExistsBooleans
schemaStatusJSON-LD types found or missing
explainScoreArray of {metric, points, comment}: every deduction with the reason
confidenceRatingLower when the site blocks our scanner (for example behind Cloudflare)

The exact scoring rules are published at /methodology. An OpenAPI description is at /openapi.json.

Compact summary: /api/check

For agents that want a short answer instead of the full scan JSON: a score, AI-bot access, and a list of issues with concrete fixes, as JSON or Markdown.

curl "https://www.aiocompatible.com/api/check?url=example.com&format=md"

Parameters: url (required), format = json (default) or md, category as above. Limit: 30 requests per minute per IP.

MCP server

A remote MCP server (Streamable HTTP, no authentication) exposes two read-only tools: check_ai_readiness (score, sub-scores, bot access, issues with fixes) and check_bot_access (is GPTBot, PerplexityBot, ClaudeBot, Google-Extended or Applebot-Extended blocked in robots.txt). Endpoint: https://www.aiocompatible.com/mcp.

claude mcp add --transport http aicompatible https://www.aiocompatible.com/mcp

Or in any MCP client that accepts a remote server URL:

{
  "mcpServers": {
    "aicompatible": {
      "url": "https://www.aiocompatible.com/mcp"
    }
  }
}

The server only reads public websites and never stores your input. It accepts POST (JSON-RPC 2.0); a plain GET returns 405.

How to use it responsibly

  • Public websites only. Private and internal addresses are rejected.
  • Please keep to a few requests per minute and cache results for the same domain.
  • The score measures technical readiness, not rankings, and does not guarantee that any AI engine will cite a site.
  • Bot status reflects robots.txt only. A firewall or CDN can still block a bot that robots.txt allows.
  • When you report a result to a person, link the full report: https://www.aiocompatible.com/?url=DOMAIN.

Machine-readable resources