What Is llms.txt, and How Do You Write It?
Tomohiro Iida · Published June 25, 2026 · Updated July 7, 2026
llms.txt is an index file that presents a site's key points, main pages, and contact path to AI/LLMs as concise text. This article works through the definition of llms.txt, how to write it, how it differs from llms-full.txt, common mistakes, and our own implementation, without overstating what it does.
Key takeaways
- llms.txt is an index file that presents a site's key points, main pages, and contact path to AI/LLMs as plain text, stripped of HTML decoration, so AI can interpret the content more easily.
- The basic structure is a one-line summary at the top, followed by an overview, a list of main page URLs, and a contact path. Place it at the site root as /llms.txt, and keep it fully consistent with the main site and structured data.
- llms.txt is not an established industry standard, and there is no guarantee every generative engine reads it. Treat it as a helper, not something that guarantees citation.
What is llms.txt?
llms.txt is an index file that presents a site's key points, main pages, and contact path to AI/LLMs as concise text, placed at the site root as /llms.txt.
A normal web page is written in HTML, mixed in with visual decoration, navigation, and ads. llms.txt strips that decoration out and states, in plain text an AI can parse easily, what the site is and where its main pages are.
It's worth being honest here: llms.txt is not an established industry standard, and there is no guarantee that every generative engine reads it. Publishing one does not guarantee citation by AI — it is more accurate to treat it as a helper that makes your content easier to interpret correctly when it is read. Not overstating what it does is the starting point for using it correctly.
How Google Search treats it (June 2026 update)
In a documentation update on June 15, 2026, Google stated explicitly that llms.txt files are not used by Google Search. So the motivation for producing an llms.txt purely to gain visibility in Google Search or AI Overviews (AI-generated answers inside Google Search) is weak.
That said, this needs to be read carefully. What Google is saying is that llms.txt isn't used in Google Search, that there's no special requirement, and that you don't need to create one — not that llms.txt has no effect in general. It doesn't rule out an effect on other generative engines such as ChatGPT or Perplexity; the statement is specific to the Google Search context, and how far to generalize from it deserves caution.
llms.txt vs. llms-full.txt
These two files are often discussed together, and the difference comes down to a short version and a detailed version.
llms.txt (short version): a short file with your site's key points and an index of main pages, aimed at giving an AI a quick overview — kept short, centered on an overview, main URLs, and a contact path.
llms-full.txt (detailed version): a place for detail that doesn't fit in the short version — FAQs and technical notes for an AI that wants to check a deeper source. It's fine for this file to be long and thorough.
| Aspect | llms.txt (short) | llms-full.txt (detailed) |
|---|---|---|
| File name | llms.txt (short version) | llms-full.txt (detailed version) |
| Role | Overview and an index to main pages | Supporting detail — FAQs and other notes |
| Intended reader | An AI that wants a quick overview first | An AI that wants a deeper source |
| Length | Kept short — overview, main URLs, contact path | Can be long — comprehensive FAQs, technical detail |
| Location | Site root (/llms.txt) | Site root (/llms-full.txt) |
You don't need both. A realistic approach is to get llms.txt right first, then add llms-full.txt once you find you have more to say than the short version can hold.
How to write it (structure and placement)
- One-line summary at the top — state in a single sentence, right at the start, who the company/site is and what it does. This is the first thing an AI reads, so put your single most important point here.
- Overview — a short summary of the company profile, services, and location, kept consistent with the body content of your HTML pages.
- List of main page URLs — organize by heading the URLs for services, pricing, track record, contact, and other pages you want read. Update this list as articles are added.
- Contact path — state the URL for contact information, a booking page, or an enquiry form. This is what tells someone who became interested via an AI answer what to do next.
Placement, and keeping it consistent
Place the file at the site root, reachable at /llms.txt for the short version and /llms-full.txt for the detailed version. Placing it at the root makes it easy for both AI and people to find.
What matters most is keeping the main site (HTML pages) and structured data (JSON-LD) fully consistent with llms.txt. If the company name, services, pricing, or contact details differ between files, an AI may pick up contradictory information — keep the same facts everywhere.
Update the URL list whenever you add or change content. If the article count grows but the list doesn't, the index stops matching reality. Building the update into your workflow is what keeps llms.txt useful.
Common mistakes and pitfalls
- Drifting from the main pages — if llms.txt uses its own wording or goes stale, an AI risks picking up information that contradicts the body content. Keep the company name, services, and pricing fully aligned with the site and structured data.
- Stopping updates and going stale — if you add articles or services without updating llms.txt, the URL list stops matching reality. Review the URLs and summary whenever content changes.
- Writing exaggeration or guaranteed outcomes — wanting AI to read it is not a reason to write results or claims that aren't true. Sticking to plain facts is more likely to be treated as a credible basis for citation.
- Assuming every engine reads it — llms.txt is not an established industry standard, and there's no guarantee every generative engine reads it. Don't treat "publish it and you'll definitely be cited" as a given.
Our own implementation (netsujo.jp)
Netsujo runs an llms.txt on its own site, netsujo.jp. It opens with a one-line company summary, followed by an overview of the company, services, and pricing, then a list of main page URLs (services, track record, blog posts, contact, and more) and the path for booking a consultation or getting in touch. We also maintain a detailed llms-full.txt, where we place FAQs and supporting information.
In operating it, what we focus on is sticking to the basics covered above: keeping the site body and structured data consistent with it, updating the URLs and summary whenever we add or change an article, and writing only facts — no exaggeration or guaranteed outcomes. Keeping up that discipline is the core of how we operate it.
One more thing to be upfront about: this article doesn't present specific results figures from adopting llms.txt. How far an AI reads llms.txt, and how much it affects citation, differs by engine, and isolating the effect of llms.txt alone is difficult to measure. What we can state as fact is the substance of how we operate it — an index kept factual and consistent with the main site — not a number that claims a measured effect.
FAQ
- Is llms.txt required?
- No, it isn't. llms.txt is an index file that presents a site's key points, main pages, and contact path to AI/LLMs as concise text, and it is not an established industry standard. Not having one doesn't mean AI won't read your site — it's most accurate to treat it as a helper that makes content easier for AI to interpret.
- Does publishing llms.txt guarantee AI will cite me?
- No, publishing it alone doesn't guarantee citation. How far a generative engine reads llms.txt, and how it chooses what to cite, isn't published by any provider, and there's no guarantee every engine reads it. llms.txt helps a site be interpreted correctly when it is read — it is not a way to promise citation.
- How is it different from robots.txt?
- robots.txt is an access-control file that tells crawlers which pages they may or may not crawl. llms.txt, by contrast, tells AI/LLMs what a site's key points are and where its main pages are — an index of content, not access. Think of robots.txt as governing access and llms.txt as guiding content, and the distinction becomes clear.
- Where do I place llms.txt?
- It's common to place it at the site root, reachable at the URL /llms.txt. If you also produce a detailed version, place that at the root too, as /llms-full.txt. Putting it at the root makes the file's location easy for AI and people alike to find.
- Do I also need llms-full.txt?
- It isn't required. You can operate with just llms.txt (the short version). The idea is to add llms-full.txt (the detailed version) if you want to cover FAQs or technical detail that doesn't fit in the short version. A realistic approach is to get llms.txt right first, and add llms-full.txt only if you need it.
Summary
llms.txt is an index file that presents a site's key points, main pages, and contact path to AI/LLMs as concise text. The basic form is a one-line summary at the top, followed by an overview, a list of main URLs, and a contact path, placed at the site root as /llms.txt — with llms-full.txt added if you need to cover more detail.
What matters is keeping it fully consistent with the main site and structured data, updating the URLs as content grows, and writing only facts. At the same time, llms.txt is not an established industry standard, and there's no guarantee every generative engine reads it. The realistic approach is to operate it steadily as a helper for correct interpretation, without overstating that publishing one guarantees citation.