Free GEO Checker (llms.txt & FAQ Schema)

llms.txt

A plain-text file at /llms.txt that starts with a title heading scores 18. A text file without a title heading scores 9. A missing file, an empty file, an error response or a web page served at that address scores 0.

Fix: Create a Markdown file at /llms.txt. Start it with a single line beginning with # and your site name, then add a short summary and a list of links to your best pages. The llms.txt checker page has the details.

AI search bots vs training bots

No robots.txt, or no AI search or assistant bot blocked, scores 18. Blocking some of the 6 search and assistant bots scores half. Blocking all 6 scores 0. Blocking only training bots, such as GPTBot or CCBot, does not lower the score.

Fix: If you only want to stay out of model training, block just the training bots and let the search and assistant bots in. See the AI crawler checker page for how blocks are read.

JavaScript-rendered content

If the raw HTML holds very little readable text (under 50 words) together with a sign that JavaScript builds the page, it scores half of 14. Otherwise it scores 14.

Fix: Serve the main content in the HTML itself, with server-side rendering or static generation, so crawlers that do not run JavaScript can read it.

Organization or Person markup

Organization or Person markup in JSON-LD with a name and a url, logo, sameAs or id scores 14. Markup without those, or without a name, scores half. No such markup scores half too.

Fix: Add Organization JSON-LD, or Person for a personal site, with your name, url and logo, so AI systems have a named entity to tie the site to.

FAQPage schema

No FAQPage markup scores 0. FAQPage markup with no questions in it scores 6 of 12. FAQPage markup with at least one question scores 12. It is found at the top level or inside an @graph.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "What does your product do?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "It helps freelancers send invoices."
    }
  }]
}
</script>

Fix: Add FAQPage JSON-LD that mirrors a question-and-answer section people can actually see on the page. Only mark up questions that are visible on the page. A minimal block looks like this:

sameAs profile links

Judged only when the page has Organization or Person markup. Links to at least one profile elsewhere score 10. With no valid profile link it scores half.

Fix: Add a sameAs list to that markup with the full URLs of your profiles, such as LinkedIn or GitHub, so the entity can be matched to the real organization or person.

Named author

Judged on pages marked as articles only. A real author name, from JSON-LD, an author meta tag or a rel="author" link, scores 8. A generic label such as admin or editor, or no author, scores half.

Fix: Name the author of each article in its JSON-LD or with an author meta tag, using a real name, so the content can be attributed to a person.

llms-full.txt

A plain-text /llms-full.txt scores 6. A missing file, an empty file or a web page served at that address scores half. A file too large to read counts as found.

Fix: This file is optional. To add it, publish the full text of your key pages as one plain-text file at /llms-full.txt, next to llms.txt.