TL;DR: llms.txt is a markdown file at your site root describing what your site is and where the important pages live. Adoption by AI providers is limited and unconfirmed. It takes five minutes, cannot hurt, and you should not expect it to move your numbers.
What it is
A proposal: publish /llms.txt, plain markdown, summarising your site and linking its key pages, so a model ingesting your site gets a clean map instead of parsing navigation chrome.
Conceptually it is robots.txt for meaning rather than permission.
# Acme
> Invoicing software for freelance designers.
## Core pages
- [Pricing](https://acme.io/pricing): Plans and what each includes
- [Features](https://acme.io/features): What the product does
- [Docs](https://acme.io/docs): Setup and API
## About
Founded 2021. Used by 4,000 independent designers.Does anything read it?
Honestly: not much, confirmed. No major provider has publicly committed to consuming it as a ranking or grounding input. Adoption is mostly among developer-tool sites.
Anyone telling you it is a major AEO lever is overselling it.
So why do it
- It costs five minutes.
- The downside is zero.
- Writing it forces a useful exercise, stating plainly what you sell, to whom, and which pages matter. Most sites cannot do this crisply, and that vagueness shows up in how models describe them.
- If adoption arrives, you are already done.
Treat it as free optionality, not strategy.
| llms.txt | robots.txt | Schema (JSON-LD) | |
|---|---|---|---|
| Purpose | Describe the site | Control crawler access | Structure page facts |
| Confirmed consumption | Limited | Universal | Widespread |
| Effort | 5 minutes | 5 minutes | Ongoing |
| Priority | Low | High | Medium |
What to do instead, first
Check your robots.txt. This is the one that genuinely matters and is broken far more often than people expect, a security plugin or a CDN default blocking GPTBot, ClaudeBot, or PerplexityBot. If your pricing page is blocked, models will keep describing your pricing as unclear no matter what else you do.
Check it in thirty seconds with the AI crawler checker, or generate a file with the llms.txt generator while you are at it.
Key takeaways
| Item | Verdict |
|---|---|
| llms.txt | Do it, expect nothing |
| robots.txt AI crawler rules | Do it first, expect real impact |
| JSON-LD schema | Worth ongoing effort |
FAQ
Will llms.txt get me into ChatGPT?
No. Nothing about it changes retrieval today.
Where does it go?
Root: https://yoursite.com/llms.txt.
Is there an llms-full.txt?
Some sites publish a full-content variant. Same verdict, harmless, unproven.