llms.txt generator and validator

An llms.txt file is plain markdown with a fixed order. Only the first part is required; the rest is what makes it useful.

1. 1An H1

   The name of the site or project, on the first line: # Ledgerly.

2. 2A blockquote summary

   One or two sentences on what you do, as a > quote under the H1.

3. 3Optional free text

   Paragraphs or lists with context, before any section. No headings here.

4. 4H2 sections of links

   ## Docs, ## Product and so on, each a list of - [Title](https://url): short note. A section named ## Optional marks links a tool can skip when it's short on space.

The validator checks each line against that structure, confirms the file is served as text rather than an HTML page, checks for an llms-full.txt, and requests up to 15 of the linked URLs to catch dead links.

We read the sitemaps listed in your robots.txt, or try /sitemap.xml and /sitemap_index.xml. For a sitemap index we follow up to three child sitemaps, preferring pages over blog posts and media. With no sitemap, we use the links on your homepage.

We take up to 200 URLs, ranked by how shallow they are (/pricing before /blog/2023/some-post), skip translated copies, and fetch the title and meta description of the top 30. Pages are grouped by their first path segment into H2 sections. Legal and careers pages go under ## Optional. Your homepage title and description become the H1 and summary.

Treat the result as a draft. The best llms.txt files are short and written by someone who knows which pages answer real questions. Cut aggressively, and if your sitemap itself looks wrong, check it with the sitemap validator.

llms.txt is a proposal, not a standard that AI search engines have committed to. Publishing one won't get you cited by ChatGPT or Perplexity on its own. What gets you cited is crawlers being allowed in and finding real content when they arrive. Check both with the AI crawler checker and What AI Sees.

Where llms.txt earns its keep today is developer documentation and APIs. Coding assistants and agent frameworks increasingly look for it, and a clean index of your docs means they fetch the right page instead of guessing.