llms.txt: What It Is And How To Write One

llms.txt is a plain text file, published at the root of your domain, that gives AI systems a curated map of your site: what it is, what it offers, and which pages matter most. It is written in simple markdown, placed at yoursite.com/llms.txt, and it exists to make your content easier for language models to find, understand, and cite correctly.

Why llms.txt exists

Language models and the retrieval systems built around them face a version of the same problem search crawlers have always faced: a website's structure is not always obvious from the outside. Navigation menus, JavaScript-rendered content, and inconsistent internal linking all make it harder for an automated system to figure out what a site actually covers.

robots.txt tells crawlers what they can access. Sitemaps tell crawlers what pages exist. Neither tells an AI system what your site is *about* in plain language, or which pages represent your best, most authoritative answers on a topic. llms.txt fills that specific gap. It is a proposed convention, not a guaranteed standard every model provider consumes, but it costs little to implement and it forces a useful exercise: deciding what your most important pages actually are.

What llms.txt looks like

The format is deliberately simple markdown. A minimal structure looks like this:

``` # Company Name

One or two sentence description of what the company does.

Docs

• Page title: short description of what this page covers

Optional

• Secondary page: short description

`

Key elements:

• An H1 with your organization or product name
• A blockquote summary describing what you do, in plain, specific language
• One or more H2 sections grouping links by purpose
• Markdown links with a short description of each linked page's content

There is no required list of section names. Group your links however makes sense for your site: services, guides, documentation, answer pages.

What to include

Be selective. llms.txt is not a sitemap dump. It is meant to be a curated shortlist of your most representative and highest-value pages. For a marketing and applied AI firm, a reasonable llms.txt might group links under headings like Services, Guides, and Answers, pointing to pages such as:

• Core service pages, for example AEO, application frontends, and AI automation
• Specific offerings under those services, like llms.txt implementation or custom CRM builds
• Direct answer pages that resolve common questions, like what is llms.txt or what is an AI agent

Each link should carry a description written the way you would describe it to a person, not SEO-stuffed keyword text. "Guide to setting up conversion tracking correctly in GA4" is useful. "GA4 conversion tracking SEO marketing analytics" is not.

How to write one, step by step

1. 01List your five to fifteen most important pages, the ones you would want an AI assistant to actually read if someone asked about your business
2. 02Group them into two or three logical sections based on purpose, such as core offerings, deep guides, and direct answers
3. 03Write a one-sentence description for each link that explains what a reader gets from that page, not what keyword it targets
4. 04Write your top-level summary blockquote in two sentences, stating plainly what your organization does and for whom
5. 05Save the file as plain markdown named llms.txt and publish it at the root of your domain, such as example.com/llms.txt
6. 06Keep it under your control in whatever system manages your site so it can be updated when your page structure changes
7. 07Revisit it whenever you launch a new pillar page or retire an old one, since a stale llms.txt pointing at dead pages is worse than none

Common mistakes

• Treating it like a sitemap. Including every page on the site defeats the purpose. Curate it the way you would curate a "start here" page for a new employee.
• Vague descriptions. A link description that just repeats the page title adds nothing. Explain what the page actually answers.
• Letting it go stale. If you restructure your SEO or design service pages and forget to update llms.txt, you are pointing AI systems at outdated paths.
• Assuming it replaces technical SEO. llms.txt is a helpful signal, not a substitute for a crawlable, fast, server-rendered site. If your content depends on heavy client-side rendering to appear, see why server-side rendering matters for SEO before worrying about llms.txt.

Does it guarantee citations?

No, and be wary of anyone who claims it does. Not every model provider currently fetches or uses llms.txt, and none has published a guarantee that its presence changes citation behavior. What it reliably does is give you a clean, low-cost artifact that documents your site's structure and intent in a format built for machine consumption. Combined with solid entity optimization and schema markup, it is one more consistent signal about who you are and what you offer, and consistency across signals is what answer engines are increasingly built to reward. You can find more general context in how do I get cited by AI assistants.

Maintaining it over time

Treat llms.txt as living documentation, not a one-time task:

• Add new pillar or answer pages as you publish them
• Remove or update links when URLs change
• Revisit descriptions periodically to keep them accurate as pages evolve
• Check it alongside your regular AI search visibility reviews rather than as a separate project

How llms.txt compares to a sitemap and robots.txt

It helps to be precise about what each of these files is actually for, since they get conflated often.

| File | Purpose | Audience | |---|---|---| | robots.txt | Controls what crawlers are allowed to access | Search and AI crawlers | | sitemap.xml | Lists every indexable URL for discovery | Search engine crawlers | | llms.txt | Curates and describes your most important pages in plain language | Language models and AI retrieval systems |

None of these three files substitutes for the others. A site can have a perfect sitemap and still lack a useful llms.txt, because a sitemap is exhaustive by design while llms.txt is meant to be selective and descriptive. Publishing one without the others leaves a gap: crawlers with no discovery path, or a discovery path with no plain-language context about what matters most.

A short example for a services business

To make the format concrete, here is a slightly longer example than the minimal skeleton shown earlier, adapted for a services company with multiple practice areas:

``` # Example Marketing Firm

Example Marketing Firm helps mid-size companies grow through search, applied AI, and server-rendered application frontends.

Services

• SEO: technical SEO, content strategy, and local search services
• AEO: visibility and citation strategy for AI answer engines
• AI Automation: agentic workflows and custom automation for internal operations

Guides

• How much does SEO cost: a plain breakdown of what drives SEO pricing
• What is llms.txt: a short definition and explanation of the format

`

Notice the descriptions are written as if explaining the page to a colleague, not as keyword strings. That is the standard to hold every line to.

Versioning and change management

Because llms.txt is just a text file, it is easy to let it drift out of sync with the rest of your site. Treat updates to it the same way you would treat updates to a sitemap: as part of your standard release process whenever page structure changes. A few practical habits that keep it accurate:

• Add a note to your page launch checklist to update llms.txt when a new pillar or answer page ships
• Review it quarterly alongside broader site audits, not just when something breaks
• Keep the file itself under version control if your site's codebase supports it, so changes are tracked like any other content change

Next step

If you are not sure which pages on your site actually deserve a spot in your llms.txt, start with an answer engine audit to see which pages are already getting attention from AI systems, then build your file around what is working plus the gaps worth filling.