How to Validate Your llms.txt File

What is llms.txt, in one paragraph?

An llms.txt file is a short, plain-text guide to your site, written for large language models and AI search tools. It lives at your-site.com/llms.txt and typically starts with an H1 title, a one-paragraph summary of what the site is, then H2 sections of links to the most important pages, each with a short description. It is an emerging convention, not a web standard and not a ranking guarantee: no file, however perfect, forces an AI system to crawl, cite or recommend your site. What a good file does is make your best pages easy to find, understand and quote accurately when an AI system does look.

Step 1: Put the file in the right place

Save the file as llms.txt at your site root — for example, https://example.com/llms.txt — not inside a subfolder and not behind a login. If your site already has one, open that URL in a browser first: a 404 here is the most common reason a validator reports nothing to check. While you are there, confirm robots.txt does not block the AI crawlers you want to reach; a perfect llms.txt behind a blocked door helps nobody. Our AI Crawler Access Checker shows which named AI crawlers can reach your site root, so check that before you fine-tune the file itself.

Step 2: Run the validator two ways

Open the validator and try the fetch mode first: enter your domain and let it retrieve the live file and score it from 0–100. Then, if you are drafting a new version, switch to paste mode and check the draft before you publish it — catching a broken link or a missing summary in a draft is cheaper than fixing it after deployment. The score is a structure and hygiene score, not a quality judgement on your content: a thin site with a tidy file can score 100, and a great site with a sloppy file can score 40. Treat the number as a to-do list, not a grade.

Step 3: Read the score check by check

• H1 title: exactly one descriptive title, naming the site and what it does. A missing or generic H1 (“Home”) tells an AI reader nothing to quote.
• Summary: one short paragraph, in plain language, saying who the site is for and what it offers. If a stranger could not explain your site from this paragraph alone, rewrite it.
• H2 sections: links grouped under clear headings (“Guides”, “Tools”, “Pricing”) instead of one long undifferentiated list. Sections are how an AI reader navigates.
• Absolute links: full URLs, not relative paths. A relative link pasted into another context points nowhere.
• Duplicates and dead links: every URL once, and every URL live. Duplicate entries waste the reader's attention budget; dead ones waste its trust.
• File size and freshness: a file that lists every page on a large site becomes noise. Prioritise durable, useful pages, and update the file when the site structure changes.

Step 4: Fix in this order

Fix missing H1 and summary first — they are what an AI summary of your site is built from. Then fix broken and duplicate links, then reorganise into sections, and only then worry about trimming length. Re-run the validator after each round of fixes; a file that scores in the 90s with accurate links beats a 100-point file that points at pages you deleted last month. Finally, keep the file in sync with the live site: add new flagship pages when they launch, remove retired ones, and re-check after any redesign or migration, when URLs are most likely to have quietly changed.

A minimal example to copy

A valid starting file needs nothing fancy — a title, a summary, and grouped links with short descriptions. Adapt this shape to your own pages, keep descriptions factual and specific, and do not pad it with pages that add nothing for a first-time reader:

# Example Tools
> Free browser tools for indie developers: formatters, checkers and launch utilities.

## Tools
- [JSON Formatter](https://example.com/json-formatter): Format and validate JSON in your browser.
- [llms.txt Validator](https://example.com/llms-txt-validator): Score an llms.txt file and fix failed checks.

## Guides
- [How to Validate llms.txt](https://example.com/how-to-validate-llms-txt): Step-by-step validation tutorial.

What validation cannot prove

A high score does not prove AI systems will crawl, index, cite or rank your site — llms.txt is a suggestion format, and each system decides for itself what to use. Keep the fundamentals correct alongside it: robots.txt, sitemaps, canonical tags, fast pages and genuinely useful content. Validation is the cheapest part of AI search readiness, which is exactly why it is worth doing first: ten minutes here removes an entire class of avoidable errors before you spend effort anywhere else.