Websites are built for browsers: navigation, scripts, cookie banners, carousels. Language models and AI agents just want the facts. llms.txt is a simple proposal for giving them exactly that.
This guide explains what it is, what it realistically does today, and how we implemented it on ficiali.com.
What is llms.txt?
llms.txt is a Markdown file served at the root of a website (/llms.txt) that gives AI systems a concise, curated summary of the site and links to the most useful content.
It was proposed by Jeremy Howard of Answer.AI in 2024 at llmstxt.org. Think of it as a cross between a sitemap.xml and a README — written for machines that read text, not for crawlers that follow links.
A typical file has:
- An H1 with the site or company name
- A short blockquote summary
- Optional context paragraphs
- H2 sections with lists of links, each with a one-line description
Many sites also publish /llms-full.txt — the full text of key pages in one file — and .md versions of individual pages.
A real example: ficiali.com
Here is a shortened version of our own llms.txt:
# FICIALI
> FICIALI is an AI-first software engineering company that builds
> AI agents, RAG systems, automation and custom web/mobile software.
## Services
- [Agentic AI & LLMs](https://ficiali.com/docs/services/agentic-ai-llm.md): AI agents and multi-agent workflows
- [RAG & Knowledge Assistants](https://ficiali.com/docs/services/rag-knowledge-assistants.md): grounded answers over your data
## Projects
- [Perach Document Intelligence](https://ficiali.com/docs/projects/perach-document-intelligence.md): AI document processing for fintech
## Company
- [About](https://ficiali.com/about.md)
- [Contact](https://ficiali.com/contact.md)
Every service, industry, project and blog post on our site has a clean Markdown twin under /docs/, so an agent can read the substance of a page without parsing HTML. This article is also available at /docs/blog/llms-txt-explained-make-your-website-ai-readable.md.
Does llms.txt help you rank in AI search?
Here's the honest answer: not directly, and not on Google.
- Google's AI search guidance from May 2026 says you don't need llms.txt or other special files for its AI features.
- No major AI search provider has published documentation saying llms.txt is a ranking or citation signal.
- Industry analyses, such as this llms.txt adoption report, suggest adoption is growing but evidence of direct visibility gains is limited.
So why bother? Because search is only one way AI reads your site:
- Agents and coding assistants increasingly fetch documentation and company pages on a user's behalf. A clean Markdown file is cheaper and more reliable for them to read than a heavy HTML page.
- Developer tools and IDE assistants commonly look for llms.txt on documentation sites.
- It forces a useful exercise — writing a crisp, factual summary of who you are and what you offer is good for every channel, including your homepage.
- The cost is tiny. On a modern framework it's a static route generated from data you already have.
Treat llms.txt as agent experience (AX) — like good API docs for your business — not as an SEO hack.
How to write a good llms.txt
- Lead with a factual one-paragraph summary. What you do, who for, where you are. No slogans.
- Link to your most important pages only. Services, products, pricing, case studies, docs, contact. Quality over completeness.
- Describe each link in one line. The description is what the model reads before deciding to follow it.
- Link to Markdown versions where you can, so the next hop is also clean.
- Keep facts consistent with your website, structured data and profiles elsewhere.
- Generate it from your data, not by hand, so it never goes stale.
How we built it (Next.js)
Our site is a statically exported Next.js app. The services, industries, projects and posts already live in typed data files, so we added route handlers that render them as Markdown at build time:
// app/llms.txt/route.ts
import { llmsTxt } from "@/lib/agent-content";
export const dynamic = "force-static";
export function GET() {
return new Response(llmsTxt(), {
headers: { "Content-Type": "text/plain; charset=utf-8" },
});
}
The same pattern produces /llms-full.txt, /services.md, /about.md and every /docs/.../*.md file. One source of truth feeds the website, the Markdown and the JSON-LD structured data.
Don't forget robots.txt
An llms.txt file is useless if the bots that would read it are blocked. Make sure the AI search and user-agent crawlers you care about are allowed — see our guide to AI crawlers and robots.txt.
The bottom line
llms.txt won't move your Google rankings, and anyone promising that is overselling it. What it does is make your business easy for AI agents to read accurately — and as more research, buying and booking happens through agents, that's worth an afternoon.
Want your website made agent-readable end-to-end? Let's talk.
