How grading works

Every grade comes from agent-landing, an open-source scanner. It fetches each homepage once, plus robots.txt, llms.txt, the sitemap, and a few well-known paths. It honors robots.txt and identifies itself in its user agent. Nobody logs in, and nothing is guessed.

Each subject gets a score from 0 to 100. The overall score is the weighted average, and the letter follows the usual school bands: A is 90 and up, B is 80, C is 70, D is 60, and anything lower is an F.

• 15%robots.txt rules for AI agentsCan the common AI crawlers and assistants fetch the homepage?
• 15%llms.txtIs there a plain map of the product written for language models?
• 15%Server-rendered HTMLDoes the page say anything before JavaScript runs?
• 10%Markdown versionsCan an agent get the page as markdown instead of layout soup?
• 10%API docsCan an agent find an OpenAPI spec or API docs?
• 10%MCPDoes the product advertise an MCP server agents can connect to?
• 10%Login wallsIs the useful content readable without signing in?
• 5%SitemapIs there a sitemap that lists the pages?
• 5%Structured dataDoes the page describe itself with JSON-LD?
• 5%Page weightHow many bytes does an agent pay to read the page?

If a site answers the scanner with a 403 or a bot challenge, the checks behind that door score zero and the report card says the robot was locked out. The site may well have an llms.txt in there. An agent that can't get in can't use it.

Who made the list

Companies that sell AI agents or agent platforms and say so on their homepage.

Regrades

Scores are reproducible. Run the same scan and you get the same answer. If you fixed your site or think a check got it wrong, open an issueand I'll rescan. The class gets rescanned every month, and who moved is the next post.