A new piece of advice about llms.txt appears every week. Let us separate what is real from what is hype.
What llms.txt is
It is a plain text file at the root of your domain (/llms.txt) meant to help a language model quickly understand what your company does and where the good information lives. It is not a mandatory standard and guarantees nothing: it is a courtesy signal, much like the sitemap once was.
What it should contain
- A headline with your brand name and one sentence describing the category.
- A short paragraph with value proposition, target audience and geographic coverage.
- A list of links to the pages you want cited: product, pricing, docs, comparisons, case studies and FAQs.
- Verifiable facts: starting prices, languages, integrations, response times.
What it should not contain
- Empty marketing copy with no data.
- Links to private or paywalled pages.
- Anything that contradicts your site: inconsistency erodes trust.
How it relates to robots.txt
robots.txt says who may enter; llms.txt says what is worth reading. If you block AI crawlers in robots.txt, your llms.txt is pointless. Decide the access policy first: blocking assistants protects your content but removes you from their answers.
What to realistically expect
Publishing it takes an hour and does no harm. But it does not replace the three levers that actually move the needle: presence in third-party comparisons, answer-shaped content and entity consistency. Treat it as a bonus, not a strategy.
How to know whether it works
The only way to check is to measure before and after: run your key question set, record the mention rate, publish the file and measure again after two and four weeks. Without that systematic measurement, any conclusion is a hunch.