Short answer: llms.txt is a markdown file that summarizes your website for AI models, published at the root of your domain (yoursite.com/llms.txt) following the llmstxt.org standard. While robots.txt tells bots "where you can go," llms.txt says "here's who I am and here's my important content." It costs nothing and makes it easier for LLMs to understand your site correctly.
What Does llms.txt Do?
Instead of crawling dozens of pages to understand a website, LLM-based engines can check llms.txt first — if it exists. A well-crafted file delivers your brand's purpose, core services or products, key pages, and contact details in a single request. This both reduces crawl budget and lowers the risk of the model drawing incorrect inferences about your brand.
How Should llms.txt Be Structured?
- H1 heading: your brand name.
- Summary block (blockquote): 2–3 sentences covering who you are, what you do, and where you operate.
- Core information: contact details, address, founding year, partnerships.
- Sections (H2): Services, Products, Guides/Blog, Corporate — each item in
[title](url): short descriptionformat. - FAQ section (optional but powerful): concise answers to critical questions about your brand — feeds the model so it describes you accurately.
- Optional section: secondary links.
Example Template
# Brand Name
> What you do in one sentence, who you serve, and what sets you apart.
**Contact:** info@example.com · +90 ...
**Founded:** 2016
## Services
- [Service A](https://example.com/service-a): one-sentence description
- [Service B](https://example.com/service-b): one-sentence description
## Guides
- [What Is X?](https://example.com/blog/what-is-x): one-sentence summary
## Featured Questions
- **What does the brand do?** Short, clear answer.
What Is the Difference Between llms.txt and llms-full.txt?
llms.txt is a summary and link map — kept intentionally small. llms-full.txt, on the other hand, presents the full text of your most important pages in a single file, allowing the model to read your site in depth without crawling it. Publishing both is best practice; adding their addresses as comment lines in robots.txt improves discoverability.
Common Mistakes
- Serving the file with incorrect encoding (non-ASCII characters break — UTF-8 is required).
- Leaving it static and never updating it: as new services or content are added, the file must be updated too (automated generation is ideal).
- Stuffing in keywords: the file should be clear enough for a human to read; models detect spam.
- Including claims that don't match your site: models cross-check against your content, and inconsistencies erode trust.
You can check whether your site has an llms.txt file — along with 21 other GEO criteria — using our free GEO audit tool. If you'd like us to handle the setup, get in touch — and you can review our own file as a live example.
