GEO · 6 MIN
What is llms.txt and does your SaaS site need one?
llms.txt is a proposed Markdown file that points AI tools to your best pages. Google says it does not use it. Here is what it is, who reads it, and when to bother.
llms.txt is a proposed convention for a Markdown file, placed at the root of a site, that gives AI tools a short, curated map of your most useful content. It is not a web standard and not a ranking factor. Google states that it does not use such files in Search and that they neither help nor harm visibility, so a SaaS site should treat llms.txt as a low-cost extra for tools that read it, not as core GEO.
- llms.txt was proposed by Jeremy Howard in September 2024 and is documented at llmstxt.org.
- The file is Markdown with an H1 title, a short summary, and sections that list links with short descriptions.
- Google says it does not use llms.txt and that the file neither helps nor harms rankings.
- Documentation platforms and some coding assistants can read these files, so docs-heavy sites get the most use from them.
- Treat it as a maintained index of your best pages, and never as a place to hide instructions.
- Nactore runs Generative Engine Optimization (GEO) as engineering work, so we add a file like this only when it serves a real reader.
What does llms.txt contain?
The proposal at llmstxt.org describes a plain Markdown file served at /llms.txt. The purpose is to give a model or an agent a concise, expert-level map so it does not have to parse a full website full of navigation, scripts, and ads. According to the specification, the only required element is an H1 with the name of the project or site. A blockquote summary is expected, and H2 sections hold lists of links with optional descriptions.
A minimal file for a fictional payments API has four parts.
| Part | Markdown element | Example content |
|---|---|---|
| Title | H1 | # Acme API |
| Summary | Blockquote | > Acme is a payments API for marketplaces. These pages cover setup, limits, and errors. |
| Section | H2 | ## Docs |
| Link entry | List item | - Rate limits: Quotas by plan |
The proposal also suggests offering clean Markdown versions of pages, typically by adding .md to a URL, so tools can fetch text without the surrounding page chrome. Some documentation platforms generate these files for you.
Does Google use llms.txt?
No. Google's guide to generative AI optimization says you do not need new machine-readable files, AI text files, markup, or Markdown to appear in Google Search, and that creating them will neither harm nor help your visibility. Its page on AI features in Search repeats that no special files or schema are required for AI Overviews and AI Mode.
That settles the Google question. For other products, vendors have not documented consistent behavior for this file, so we cannot claim that a given assistant fetches it, weights it, or follows it. If someone sells llms.txt as a way to rank in AI answers, ask for the vendor documentation. In our reading, none exists for ranking.
Who might actually read it?
The honest answer is tools that are pointed at your site on purpose. Here is where it plausibly helps.
| Reader | Likely use | Evidence level |
|---|---|---|
| Developers using coding assistants | Paste or fetch a docs index to give the assistant context | Practical use by developers, vendor-neutral |
| Agents with a browsing tool | Find the right docs page without crawling everything | Depends on the agent, not guaranteed |
| Docs platforms | Auto-generate and serve the file | Documented by those platforms |
| Google Search | None | Google says it is not used |
| Chat assistants answering general questions | Unknown | Not documented |
This is why docs-heavy products are the best candidates. If your customers integrate through code and use assistants while they build, a clean docs index has a real audience. A marketing site with twelve pages has little to gain.
How do you write a good one?
If you decide to publish a file, keep it small and honest.
- Name the product and say what it does. One H1 and one blockquote, in plain sentences.
- List your best 10 to 40 pages. Quickstart, core guides, limits, errors, security, and the changelog.
- Describe each link in a few words. Say what the reader will get, not what you want them to feel.
- Group by task. Use H2 sections such as Getting started, Guides, Reference.
- Put lower-value pages under Optional. The proposal uses that section name for content that can be skipped.
- Keep it true. Remove dead links and renamed pages whenever the docs change.
- Do not hide instructions. Text aimed at steering a model rather than describing your site is manipulation and may be treated as such.
Treat the file like any other shipped asset. Add a check to CI that verifies every listed URL returns a successful response, as you would for a sitemap.
Generate the file from the same source as your docs navigation. A hand-written index will drift within a month, and a stale map is worse than none.
What matters more than llms.txt?
Almost everything on the standard list. Before you spend an afternoon on this file, check these.
- Crawl access. Public pages should not be blocked by robots rules, a firewall, or a login. OpenAI, Anthropic, and Perplexity publish their bot names in the OpenAI bots page, Anthropic's guidance, and Perplexity's list.
- Indexing. Sitemaps, canonicals, and no accidental noindex.
- Real answers. Pages that state the solution in the first paragraph. See how to write content AI answer engines can cite.
- Accurate facts. One consistent description across your site and profiles.
- Measurement. A prompt panel that shows what assistants actually say, as in tracking AI citations for SaaS.
The order of priority is the same one laid out in what is GEO. The file sits at the end of the list, after the work that every system depends on.
Does llms.txt create risk?
Mostly small ones. A file that lists private or unreleased URLs can leak them. An outdated file can send tools to deprecated docs. And a file may create a false sense of progress, where a team ships it and stops looking at the pages behind it. None of these are reasons to avoid the file, only reasons to maintain it.
Frequently asked questions
Is llms.txt an official standard?
No. It is a proposal published at llmstxt.org and adopted by some documentation platforms. No standards body has ratified it, and AI vendors have not committed to one behavior.
Will llms.txt get my site cited by ChatGPT or Claude?
We cannot say that, and no vendor documentation supports it. Citation depends on retrieval, relevance, and trust in the content. The file may help tools that are deliberately pointed at it.
What is the difference between llms.txt and robots.txt?
Robots.txt tells crawlers what they may fetch. llms.txt is a curated reading list for tools that want a summary of your key pages. One restricts access and the other suggests content.
How much effort should we spend on it?
An hour for a generated file on a docs site is reasonable. Do not delay fixing crawl access, indexing, or content quality to work on it.
A small file, kept honest
Publish llms.txt if your audience builds with assistants, generate it from your docs, and keep working on the pages themselves. Want this built for your team? Book a free 30-minute call.
Want to apply this to your business?
Book a free 30-minute call. We will tell you what we would do first.