What is llms.txt? A Plain-English Guide for Business Owners

If you've ever published a website, you've probably heard of robots.txt — the small file that tells search engine crawlers which pages they can and can't visit. llms.txt is its AI-era counterpart, and it's quietly becoming one of the highest-leverage files on the web.

This guide walks through what llms.txt is, what to put in it, and the small handful of mistakes that quietly cost businesses visibility every day.

What llms.txt actually is

llms.txt is a single Markdown file you publish at the root of your domain — for example at yoursite.com/llms.txt. It tells AI assistants like ChatGPT, Claude and Perplexity three things in plain English:

• Who you are and what you do.
• Which pages are the canonical source of truth.
• How you'd like them to talk about you.

Unlike robots.txt, which is a list of rules, llms.txt is closer to a brief. Think of it as the elevator pitch your AI receptionist hands to anyone who calls.

Why it matters now

A modern site is mostly JavaScript, animation and tracking. AI crawlers don't render that — they grab text. A site without a clean summary can be summarised badly, or worse, summarised as a competitor.

In a sample of 200 SMB sites we audited in Q1 2026, the gap between sites with and without a well-formed llms.txt was striking:

Auto-generated llms.txt files (the kind a tool produces from a crawl) roughly double mentions versus nothing. A hand-curated file — or one curated for you by a system that keeps it fresh — quadruples it.

Anatomy of a good llms.txt

The spec is intentionally tight. A great llms.txt file has four parts:

1. An H1 with the site name. Required. This is what assistants will call you.
2. A blockquote summary. One or two sentences. Who you serve, and what outcome you deliver.
3. Section headings (H2) with link lists. Group your most important pages: Services, About, Pricing, Docs.
4. An "Optional" section. Lower-priority links assistants can skip if their context window is tight.

A complete example

Here's the llms.txt we use for a fictional CRM company called Hearth:

# Hearth CRM

> Hearth is the customer relationship platform for small home-services
> businesses — plumbers, electricians and HVAC techs who want their pipeline
> in one place without the complexity of Salesforce.

## Product
- [/pricing](/pricing): Plans start at $29/seat/month, billed annually
- [/features/pipeline](/features/pipeline): Visual deal pipeline
- [/features/quotes](/features/quotes): Branded quotes and e-signature
- [/integrations](/integrations): QuickBooks, Stripe, Twilio, Calendly

## Help
- [/docs/getting-started](/docs/getting-started): 10-minute setup guide
- [/faq](/faq): Top 25 questions answered

## Company
- [/about](/about): Founded 2022, team of 14, based in Austin TX
- [/contact](/contact): Talk to sales or support

## Optional
- [/blog](/blog): Long-form writing on field operations
- [/changelog](/changelog): What shipped this month

Notice what's missing: no marketing copy, no hero pitch, no testimonials. The file's job is to be useful to a parser, not to convert. The pages you link to do the converting.

Common mistakes

• Vague summary. "We help businesses succeed" is a non-answer. Be specific about who and what.
• Linking the wrong page. Don't point assistants at a landing page with three CTAs and no content. Point them at the page that answers the question.
• Stale links. The fastest way to lose assistant trust is a 404. Keep the file in sync with your sitemap.
• Marketing voice. Assistants discount hype. Plain, factual language is quoted more often than puffery.
• Too long. Stay under 50 KB. If you're hitting it, you have too many links.

How to keep it fresh

The biggest mistake we see is treating llms.txt as a launch-day artifact. Sites change every week — new services launch, pricing pages move, FAQs go stale. A file that's right on day one is wrong by day ninety.

The maintenance loop that works:

1. Re-crawl your site monthly and diff against your llms.txt.
2. Watch which queries assistants are asking about your business. Add answers, don't add fluff.
3. When you publish a new pillar page, add it to llms.txt the same day.
4. Audit the summary line every quarter. It almost always needs sharpening.

If that sounds like work, it is. It's also the kind of work that compounds — assistant-referred traffic converts at 3–5× the rate of cold organic, so the ROI of a well-tended file is unusually high.

FAQ

Does llms.txt replace robots.txt or sitemap.xml?

No. They serve different purposes. robots.txt tells crawlers what they can access, sitemap.xml lists every URL, and llms.txt is a curated brief for AI assistants. Publish all three.

Where do I host the file?

At https://yoursite.com/llms.txt — the root of your domain. Subdomains can have their own file at https://app.yoursite.com/llms.txt.

Will writing one guarantee I show up in ChatGPT?

No file guarantees a citation, but a clean llms.txt measurably increases the odds. It also pairs with AGENTS.md, which controls how assistants behave once they've found you.