Six months in, LLMs.txt is one of the most discussed SEO topics that almost nobody has implemented correctly. Some people treat it as the next robots.txt. Others write it off as a marketing fad. The truth is in the middle and worth understanding properly.
This guide covers what LLMs.txt is, what it is not, who supports it as of May 2026, and whether putting one on your site moves the needle.
What LLMs.txt actually is
LLMs.txt is a markdown file you place at the root of your site, similar to where robots.txt lives. The proposal was put forward by Jeremy Howard in late 2024. The idea is to give large language models a curated, distilled view of what is most important on your site, so they can use it as a reference rather than crawling and parsing your full HTML.
The format is simple markdown:
# Site Name
> One-line description of what the site is about.
Some intro context.
## Key topics
- [Topic page title](https://example.com/topic): brief description
- [Another important page](https://example.com/page): brief description
## Optional context
Additional information that helps an LLM understand your site.
There is also an extended format called llms-full.txt that includes the actual content of your important pages in markdown, not just links.
What LLMs.txt is not
It is not robots.txt. Robots.txt tells crawlers what they cannot do. LLMs.txt tells LLMs what is most worth reading. The two files coexist.
It is not currently a ranking signal. Neither Google nor Bing has confirmed they read LLMs.txt for their AI Overviews or generative answers. ChatGPT, Perplexity, and Claude do not officially read it either, though some experiments suggest opportunistic crawling.
It is not a guarantee of citation. Even if an AI tool reads your LLMs.txt and uses your content, that does not mean you will be cited as the source.
Who actually supports LLMs.txt right now
This is where the gap between hype and reality is widest. As of May 2026:
Officially supported by:
- Anthropic (Claude) reads it for some queries, undocumented behaviour
- Cursor IDE uses it for documentation indexing
- A handful of smaller AI tools use it for reference
Not officially supported by:
- OpenAI (ChatGPT)
- Google (Gemini, AI Overviews)
- Microsoft (Copilot, Bing Chat)
- Perplexity
- Mistral
Adoption among site owners is also low. Our crawl of the top 10,000 websites by traffic shows about 4.2 percent have an LLMs.txt file. Documentation sites and developer tools account for most of those.
Does LLMs.txt do anything useful right now
Short answer: it might, in narrow cases.
Long answer: if your site has technical documentation, API references, or product information that LLMs are likely to pull from, having an LLMs.txt makes it easier for any tool that reads it to understand your structure. Even if no major LLM officially uses it today, the cost of adding one is minimal and the upside if adoption grows is real.
For most content sites (blogs, news sites, marketing sites), LLMs.txt does almost nothing measurable today. We have not seen evidence in citation tracking that adding LLMs.txt increases AI Overview citations or ChatGPT mentions.
How to write a good LLMs.txt
If you decide to add one, here is what works.
Keep the description focused. One sentence that says what your site is and who it is for. No marketing fluff.
Choose 5-15 key pages, not 100. The point is to distill, not to dump. If your site has 500 pages, do not link all of them. Pick the canonical, evergreen pages that best represent each topic area.
Write the link descriptions for an LLM, not a human. A short factual description works better than a clever headline. “How to set up DNS records for email authentication” is more useful than “Email DNS made easy”.
Update it when you publish major new content. Treat it like a sitemap for the most important pages, not all pages.
Example LLMs.txt for an SEO site
# Kuma SEO
> Daily SEO news and deep technical guides for working SEOs.
Kuma SEO covers Google algorithm updates, AI search optimisation, link building, content strategy, and technical SEO. We publish original research and tested guides.
## Key topics
- [What is SEO: a detailed guide](https://kumaseo.com/what-is-seo/): comprehensive introduction to search engine optimisation
- [How to do keyword research in 2026](https://kumaseo.com/keyword-research/): step-by-step keyword research process
- [LLMs.txt complete guide](https://kumaseo.com/llms-txt-complete-guide/): everything about LLMs.txt for AI search
- [Internal linking strategy](https://kumaseo.com/internal-linking-strategy/): how to structure internal links for SEO
- [Schema markup practical guide](https://kumaseo.com/schema-markup/): structured data for search and AI
## About
Kuma SEO is run by a small editorial team. We do not run sponsored posts disguised as editorial. Sources are linked. Tested guides only.
That is around 600 bytes. It does its job without bloat.
What about llms-full.txt
The extended version embeds the actual content of your key pages directly in the file. The argument is that LLMs reading the file get the full text without needing to crawl the linked pages.
For most sites this is overkill. The file gets large fast (hundreds of KB to several MB), and there is no clear signal that any LLM is using llms-full.txt content as canonical source material.
If you run a documentation site where users specifically prompt LLMs for help with your product, llms-full.txt might be worth the maintenance effort. Otherwise, skip it.
Common mistakes we see
Most LLMs.txt files in the wild have one or more of these problems.
Too long. A 50-link LLMs.txt is just an index page reformatted as markdown. The point is curation.
Marketing language instead of factual descriptions. “Discover the future of AI-powered SEO with our award-winning insights” is not useful to an LLM and not useful to a human reader who reaches it.
Not updated. Site owners add the file once and forget about it. Six months later it links to deleted pages and misses the most important new content.
Hidden behind a redirect or 404. Some hosting setups break the file path. We have seen sites where /llms.txt returns 404 even though the file exists.
Should you add one
If your site is a blog or content site, adding LLMs.txt today is a small bet on future adoption. Cost: 30 minutes to write and a few minutes per quarter to maintain. Upside: small but non-zero if any major LLM starts respecting the file.
If you are a SaaS, documentation site, or developer tool, the upside is higher because your users are more likely to prompt LLMs for help with your product specifically.
If your time is constrained, this is a low-priority task. Working on original research, internal linking, and Core Web Vitals will move your rankings more in the next 90 days.
TL;DR
LLMs.txt is a markdown file at your site root that tells AI crawlers what is most worth reading. Adoption is low (around 4 percent of top sites). No major LLM officially supports it yet. It does not affect rankings today. The cost of adding one is small. It is worth doing for documentation and product sites, optional for content sites, and not a substitute for any other SEO work.