Every score on this site comes from a fixed set of public checks against a dealership's website. Nothing here is opinion, and nothing is graded on a curve. A 100 means every check passed. A 40 means most of the agent layer is missing.
What we fetch
The homepage, robots.txt, llms.txt, sitemap.xml, and a short list of well-known paths for AI discovery, MCP, and inventory endpoints. Requests use a standard desktop browser user agent. Sites that refuse the request are marked blocked and rescanned through residential connections on the next cycle rather than scored.
The checks and their weights
·
MCP endpoint
An MCP server lets AI agents query the store directly. This is the biggest single gap on nearly every dealer site.
20
·
Machine-readable inventory
A JSON inventory feed at a known path. Without it, AI has to guess what is on the lot.
15
·
llms.txt
A plain-text file at /llms.txt that tells language models what the site is and where the important pages are.
10
·
AI discovery file
A machine-readable file describing the business for AI agents.
10
·
AutoDealer schema
JSON-LD that identifies the site as a car dealership, with address, hours, and brands.
10
·
Any structured data
At least one JSON-LD block on the homepage.
4
·
Organization or LocalBusiness schema
Entity-level schema that connects the store to its parent and its location.
3
·
FAQ schema
Marked-up questions and answers that AI can quote directly.
3
·
GPTBot allowed
robots.txt does not block OpenAI's crawler.
5
·
ClaudeBot allowed
robots.txt does not block Anthropic's crawler.
5
·
Google-Extended allowed
robots.txt does not block Google's AI training crawler.
5
·
XML sitemap
A sitemap declared in robots.txt or served at /sitemap.xml.
5
·
Name and phone on page
The dealer's name and phone number appear in the homepage HTML, so AI can match the site to the business listing.
5
Total: 100. The agent layer (MCP, inventory, llms.txt, AI discovery) carries 55 points because it is what lets an AI agent act, not just read. Schema carries 20, crawler access 15, sitemap and business identity 10.
What this does not measure
Whether ChatGPT, Gemini, Google AI Mode, or Grok actually recommend the store. That is a separate, prompt-based measurement that depends on engine, location, and time, and it is what PromptGraph tracks for its clients. Readiness is the cause. Visibility is the effect.
Corrections
Every check is reproducible with a browser. If a result is wrong, email Gray@PromptGraph.ai with the store and the check and we will rescan within two business days.