How Lensy scores a page.

Give Lensy a URL, and it shows you how AI search engines and coding agents read your page, and what to fix first.

The process

  1. Paste a URL

    Enter one docs page: an API reference, a guide or a tutorial. Lensy checks that page, not your whole site.

  2. Lensy reads the page

    It fetches the page as an honest bot, under its own name, along with the files bots read on your site: robots.txt, the sitemap, llms.txt and the page's Markdown version.

  3. You get two scores

    One score for coding agents and one for AI search, each out of 100, with the fixes that would add the most points first.

Start a free scan

What Lensy fetches

  • The page's HTML

    As an honest bot that names itself and doesn't run JavaScript, the way AI crawlers and most coding agents read it.

  • The page as Markdown

    The same address asked for with Accept: text/markdown, and its .md address, which agents prefer to HTML.

  • robots.txt

    To see which AI search and retrieval bots your rules allow on this page.

  • Your sitemap

    To see whether the page is listed, so search engines can find it.

  • llms.txt

    The index coding agents look for first, to see whether it exists, resolves and links to Markdown.

  • A missing page

    An address that doesn't exist on your site, to see whether it returns a real 404 rather than a page that looks like content.

The two scores.

Coding agents measures how well a tool such as Claude Code, Cursor or Copilot can find and read the page. It's built on AFDocs, the open Agent-Friendly Documentation checks: an llms.txt index, Markdown versions of pages, real 404s for missing pages and content that loads without JavaScript. Lensy adds one check of its own, for headings an agent can navigate. Each check has a weight (critical 10, high 7, medium 4, low 2), and the score is the weight earned over the weight of the checks that ran. With no llms.txt the score is at most 59. If AFDocs can't run on a page, six Lensy checks (C1 to C6) give the score instead.

AI search measures how well engines such as ChatGPT search, Perplexity and Claude can retrieve the page and cite it. It has 6 checks worth 100 points between them.

Content that loads without JavaScript counts in both scores, because both kinds of tool read the raw HTML.

Above both scores, a reachability line says whether an honest AI agent could load the page at all.

Every check has its own page with its points, what it measures and the evidence behind it, and the research behind the checks is in Learn.

A local build.

The command line tool can also check a docs build on your machine, before you publish it. Some checks depend on your web server or CDN rather than your pages, such as Markdown on request, real 404s, redirects and cache headers, so a local build can't test them. Lensy leaves those out and shows the score as the points it could check, for example 71 of 84 points checked, never stretched to 100.