What is llms.txt?
llms.txt is a proposed plain-text file, written in Markdown and placed at the root of a website, that summarises the site for large language models and links to its most useful pages. Jeremy Howard of Answer.AI proposed it in September 2024. The idea is that a model, or a tool working for a person, can read one short, clean file instead of working through menus, scripts and layout to find what matters.
It sits beside robots.txt and sitemap.xml but does a different job. robots.txt says what crawlers may fetch, a sitemap lists the pages a site wants indexed, and llms.txt describes what the site is and where the best answers are.
What goes in the file?
The proposal sets out a simple structure:
- a top-level heading with the name of the business or project, which is the only required part
- a short summary in a quoted block saying what the business is and does
- optional paragraphs with the details a reader needs to use the rest of the file
- sections under second-level headings, each a list of links with a one-line note on what each page answers
- an optional section for secondary pages that can be skipped when space is short
Some sites also publish a longer llms-full.txt containing the full text of their key pages. That convention grew up around the proposal rather than being part of the original specification.
Do Google or the AI assistants use it?
Not in any way that has been confirmed. Google's search team has said publicly that Google Search does not use llms.txt, and John Mueller of Google compared it to the keywords meta tag, a signal search engines stopped trusting long ago. None of the major AI companies has said that its assistants read llms.txt when deciding which business to mention or which page to cite.
Where llms.txt is used today is closer to the developer world. Some coding assistants and documentation tools read it when a person points them at a site, and some documentation platforms generate it automatically. That is a real use, but it is not the same as ranking in search or being recommended in an AI answer.
Then why does CompanyFirst publish one?
Because it is cheap to do properly and does no harm. Our llms.txt and llms-full.txt are generated from the same content files as the pages themselves, so they list every service, industry page, case study and article with the same titles and descriptions, and they can never drift out of date. We did the same for Floret Resorts, whose llms.txt lists every page with its own title and description.
We do not count it as a ranking factor, and we would not recommend paying much for one. If a website's platform can generate it automatically from content that is already accurate, it is worth switching on. If someone is selling it as the key to AI visibility, it is not.
What matters more than llms.txt?
Almost everything else on the list. The things that decide whether an assistant can find and describe a business are older and more ordinary:
- search and AI crawlers can reach the site, with nothing blocking them in robots.txt, the firewall or the CDN
- every important offer has its own crawlable page that answers the questions customers ask
- the business's name, address, phone and key facts match on the website, Google Business Profile and listings
- structured data describes the business and its offers, and matches the visible page
- the site is in Google's and Bing's indexes, with a sitemap submitted to both
A site that gets these right will be readable by AI systems whether or not it has an llms.txt. A site that gets them wrong will not be rescued by one.
Does a small business need one?
No, not as a priority. A hotel, travel agency or property developer should fix crawler access, offer pages, verified facts and listings first. Once those are in place, an llms.txt is a reasonable finishing touch: a few minutes' work when generated from the site's content, and a tidy summary for any tool that does read it.
Where this fits in the audit
The Digital Footprint Audit treats llms.txt as part of Layer 09, reach and amplification, and checks it only after the layers below it. The audit records whether the file exists and whether it agrees with the site, but it never ranks it above crawl access, verified facts or offer pages.