Back to blog
#seo#ai#web-development

llms.txt in 2026: Get Your Content Cited by AI

llms.txt is a markdown map that points AI crawlers at your best pages. Here is what it does, what it honestly does not, and how to ship one that helps.

By Rafael Costa4 min readEnglish
Share
llms.txt in 2026: Get Your Content Cited by AI

A growing share of your buyers now ask ChatGPT, Perplexity or Google's AI Overviews before they ever open a search results page. Those systems do not read your site the way a person does. They send a crawler, strip the menus and scripts, and try to work out which pages actually answer the question. llms.txt is a small file that tries to make that job easier by handing the crawler a clean, hand-ordered map of your most important content in plain markdown.

It has become one of the most over-promised ideas in technical SEO. Vendors sell it as a switch that buys you citations in AI answers. It is not that. As of mid-2026 no major AI provider has publicly committed to reading it in production, and it is not a ranking or citation factor. What it is, is cheap, harmless future-proofing that takes under an hour and pairs well with the structured-data work you should already be doing. This guide separates what llms.txt genuinely does from the hype, and shows how to ship one that earns its place.

What llms.txt actually is

The file lives at the root of your domain, https://yourdomain.com/llms.txt, and it is just markdown. A top-level # heading names the site, an optional blockquote summarizes what you do, and then ## sections group links to your key pages with a short note on each. That is the whole format. No new syntax, no schema to learn.

The idea borrows from robots.txt and sitemap.xml but solves a different problem. A sitemap lists every URL for a search crawler to index. llms.txt does the opposite: it curates a short list of the pages that best explain your product, your pricing and your expertise, so a model retrieving context does not have to guess. Think of it as the table of contents you would hand a new hire who has one hour to understand the business.

What it does, and what it honestly does not

Be clear-eyed here, because most articles on this topic are not. llms.txt does not make a model cite you, does not affect Google rankings, and is not yet read by OpenAI, Anthropic, Google or Mistral in their production answer systems. Anyone promising citations in exchange for the file is overselling.

What it can do is more modest and still worth it. It gives any crawler that chooses to support it a clean retrieval map instead of forcing it to parse a JavaScript-heavy page. It is a low-effort signal that costs nothing to maintain and carries no downside if ignored. And it forces a useful exercise: deciding which ten or fifteen pages truly represent you.

Treat it as insurance, not a growth lever

If a tool or agency frames llms.txt as the thing that will get you into AI answers, be skeptical. The real work that earns citations is clear, answer-first content and solid structured data. The file is a cheap complement to that, not a substitute.

How to write one that earns its place

A good llms.txt is short and curated. Resist the urge to dump your sitemap into it; a list of 200 links is noise, and noise is the opposite of the point. Aim for the pages a smart stranger would need to understand and trust you.

markdown
# Lusivision

> Premium custom-software studio in Portugal building web, mobile and
> AI products for businesses worldwide.

## Core pages
- [Services](https://lusivision.com/en/services): web, mobile, cloud and AI consulting
- [Portfolio](https://lusivision.com/en/portfolio): selected client work
- [Contact](https://lusivision.com/en/contact): start a project

## Key guides
- [What an AI agent costs to build](https://lusivision.com/en/blog/ai-agent-development-cost-2026)
- [Generative Engine Optimization](https://lusivision.com/en/blog/generative-engine-optimization-ai-search)

Lead each link with the destination, then a few words on what it covers, because that one-line description is the signal a model reads. Keep the descriptions factual and specific. Some teams also publish an llms-full.txt that inlines the actual content of those pages as markdown, which spares a crawler the round trips; only bother if your key pages are stable and you can keep it current.

The hard rule is freshness. A map that points at retired pages or stale pricing is worse than no map. If you cannot commit to updating it when your site changes, do not ship it.

Where it sits next to robots.txt and structured data

llms.txt does not replace anything. Keep robots.txt for crawl rules and sitemap.xml for full indexing; the new file is additive. The work that actually moves the needle for AI visibility lives elsewhere, and llms.txt only helps once that foundation is in place.

  • Server-rendered HTML. If your answer only appears after a client-side fetch, many crawlers never see it. Ship content in the first response. Our Next.js SEO playbook covers the how.
  • JSON-LD on every important page. Article, Organization and FAQPage schema give models labeled facts instead of forcing them to infer structure.
  • Answer-first content. This is the real lever. Our guide to getting cited by AI goes deep on writing passages a model can lift cleanly.

Should you ship one in 2026?

For most sites, yes, with the right expectations. It takes under an hour, several CMS plugins generate it automatically, and it carries no risk if AI platforms ignore it. The cost is trivial and the option value is real: if adoption grows, you are already positioned, and if it does not, you have lost almost nothing.

What you should not do is treat it as a strategy. The brands that get cited in AI answers earn it by publishing clear, credible, machine-readable content, not by adding a file. Put the file in place, then spend your real effort on the content behind it. If you want help making your site legible to both search engines and AI, tell us what you are building.

#seo#ai#web-development
Share this article
Rafael Costa

Written by

Rafael Costa

Software Engineer & Technical Writer

Rafael is a software engineer at Lusivision who writes about web development, cloud architecture and applied AI. He has spent over a decade shipping production software for companies across Europe and enjoys turning hard technical topics into clear, practical guides.

View all articles

Related articles

Migrating to Next.js 16: A Practical Upgrade Guide
EN
#web-development#performance

Migrating to Next.js 16: A Practical Upgrade Guide

Next.js 16 makes caching opt-in, turns params into promises, and swaps middleware for proxy.ts. Here is how to upgrade an App Router app without breaking it.

5 min read

Newsletter

Stay in the loop

Occasional notes on software, design and what we're building. No spam — unsubscribe anytime.