llms.txt is a proposed file that sits at the root of your site. It hands AI systems a short, curated map of your most important pages, written in plain markdown. No major assistant has publicly confirmed that it reads the file, and it is not an official standard. Treat it as low-cost future-proofing, not a ranking lever.
What llms.txt actually is
llms.txt is a single plain-text file, written in markdown, that you place at the root of your domain (for example, yoursite.com/llms.txt). It is meant to be read by large language models and the systems built on them. The idea is simple: give a model a clean, curated summary of what your site is and where its best content lives, instead of making it guess from raw HTML.
A typical file starts with a heading that names your company or product, followed by a one or two line summary of what you do. Below that, you group links under short markdown headings, and each link gets a brief description of what the reader will find there. The goal is a tidy table of contents, not a full copy of your site.
There is also a fuller variant, usually called llms-full.txt, that expands the same idea. Instead of pointing to pages, it inlines the actual content in one long markdown file, so a model can read your key material without following links. You can publish one or both. The short file is a map to your content. The full file carries the content itself.
How it differs from robots.txt and sitemap.xml
People often lump llms.txt together with robots.txt and sitemap.xml because all three are small files at the root of a site. They do very different jobs, and llms.txt does not replace either of the others.
robots.txt is about permission. It tells crawlers, including AI crawlers, which paths they may or may not fetch. It is a set of rules, not a recommendation of what is worth reading.
sitemap.xml is an inventory. It lists your URLs in a machine-readable format so crawlers can discover every page, including ones that are hard to reach by following links. It aims for completeness, not curation.
llms.txt is a guide. It is written for a model's benefit, and it is curated by a human. Rather than listing every URL or setting access rules, it points to the handful of pages you most want understood and explains, in one line each, why they matter. Permissions, inventory, and guidance are three separate jobs, so the three files sit side by side.
The honest status
This is the part that matters most. llms.txt is an emerging convention, not an official standard. It was proposed by the community, and adoption so far is voluntary and uneven, concentrated among documentation sites and developer tools.
The major AI assistants have not publicly confirmed that they read llms.txt or give it any weight. That does not mean none of them do. It means you should not assume the file changes how any specific engine answers, and you should be skeptical of anyone who promises that it will.
So why publish one at all? Because the cost is low and the downside is close to zero. Writing a clean, accurate summary of your best content is a useful exercise on its own, and if support grows, you are already prepared. Think of it as cheap insurance and good hygiene, not a growth tactic.
The honest framing is this: llms.txt is worth doing because it is easy and harmless, not because it is proven to move visibility. Your real levers stay the same, clear content, good structure, and being cited by sources that models already trust.
How to write a good one
Lead with identity. The first line should name you and say plainly what you are, in the words a customer would use. A model reading the file should understand your category within the first sentence.
Then link your anchors. Point to your canonical explainer pages, your core product pages, your documentation, and your pricing, and give each a short, factual description. Curate hard. A focused file with ten strong links is more useful than a dump of a hundred.
Keep it accurate and current. Because the file is meant to represent your best content, a stale or misleading llms.txt is worse than none. When you change a key page or launch a new one, update the file. Treat it as living documentation, not a one-time task.
Do not overpromise, in the file or to yourself. llms.txt cannot force a model to cite you, and stuffing it with keywords or claims will not help. Write it as if a careful human editor were the reader, because in effect the model is standing in for one.