GEO
llms.txt: how to guide AI on your website
llms.txt explained in plain terms: what this file really does for GEO, how to write it yourself in minutes, and what it doesn't replace (SEO, robots.txt).
A vendor or an SEO plugin mentioned a file called "llms.txt" that you should add to your website. Direct answer: it's a small text file, placed at the root of your site, meant to summarize its content for artificial intelligence systems — but it's neither an official standard nor an SEO lever, and most AI systems don't read it yet. Here's what it actually does, how to write it in ten minutes, and why it should never replace the real work on your content.
| Benchmark | Value |
|---|---|
| Date of the initial proposal | September 3, 2024, by Jeremy Howard |
| llms.txt files never actually requested | 97% of valid files analyzed (Ahrefs study, May 2026) |
| Google Search's official position | The file neither helps nor hurts rankings — Google ignores it |
The essentials
- llms.txt is not an adopted standard: it's a 2024 proposal, not officially recognized by Google, OpenAI, or Anthropic.
- It blocks nothing and allows nothing: unlike robots.txt, it contains no access rules, only descriptive links.
- It doesn't improve your Google ranking: Google Search's official documentation is clear on this point.
- In practice, almost nobody reads it: a recent study puts the share of files that receive zero requests at 97%.
- Creating it stays low-cost: ten minutes of work, no risk — but it's not where your AI visibility is actually decided.
What is the llms.txt file?
The llms.txt file is a Markdown document placed at yourdomain.com/llms.txt, meant to give artificial intelligence systems a structured summary of a site's important pages. The idea was proposed on September 3, 2024 by Jeremy Howard, founder of Answer.AI, as a more readable alternative to a classic XML sitemap.
Its structure follows a precise syntax. An H1 heading carries the site or project name — the only mandatory section. A short summary in blockquote form follows, then optional free text. The core of the file consists of H2-level sections containing lists of links, in the format [page name](url): descriptive note. A final section named "Optional" flags secondary content that the AI can skip if it's short on context.
A version 2 of the proposal has since been published, last updated on August 10, 2026. It adds the possibility of per-subfolder llms.txt files and Markdown versions of each page, linked via dedicated technical headers. The underlying principle echoes schema.org structured data: helping a machine understand content, not a human reading the page.
llms.txt, robots.txt, sitemap.xml: don't confuse them
Three files coexist at the root of a site, and they don't play the same role. Robots.txt gives crawling robots instructions to block or allow access. Sitemap.xml exhaustively lists every page to help search engines discover them. Llms.txt does neither: it's a curated selection, designed to point an AI toward what matters most.
| File | Role | Access control? |
|---|---|---|
| robots.txt | Allow or block page crawling | Yes |
| sitemap.xml | List every page for indexing | No |
| llms.txt | Summarize and prioritize content for an AI | No |
The classic mistake: believing that llms.txt can, like robots.txt, prevent an AI from reading certain pages. It can't. The official specification defines no "allow" or "block" rule — only lists of links. If you want to exclude content from crawling, that's robots.txt's job, not llms.txt's.
Does llms.txt actually do anything today?
So far, the real-world usefulness of llms.txt remains unproven, and the most authoritative voices on the subject are rather skeptical. John Mueller, from Google's Search team, summed up the situation on Bluesky on June 17, 2025: "FWIW no AI system currently uses llms.txt" (Search Engine Roundtable). He added that mainstream conversational agents do fetch a site's pages, but never this particular file, based on server log analysis.
An Ahrefs study covering 137,210 domains measured the same phenomenon from the server side in May 2026: 28% of the sites analyzed have a valid llms.txt file, but 97% of those files received absolutely no requests that month (Ahrefs). Among the rare files that were accessed, bots accounted for 96% of the traffic — and AI bots, across all categories, made up only 19.5% of those requests.
Google's official position: Google Search Central's documentation states that you don't need to create an llms.txt file to appear in Google Search, and that doing so "neither helps nor hurts" a site's visibility, since Google Search ignores it.
So why does everyone keep talking about it? The file is cheap to produce, and some specialized agents — developer tools, assistants that actively navigate a site to complete a specific task — may use it once they're already on the site, even though that's not the norm for information search. It's a low-stakes bet, not a visibility strategy: if the real goal is to be recommended by a generative AI, it's worth first understanding what GEO changes compared to classic SEO.
How to create your own llms.txt file
No paid tool is needed: a text editor is enough. The structure to follow fits in a few lines.
# Your Company Name
> One-sentence summary of your business and what the site offers.
## Main pages
- [Page name](https://yoursite.com/page): short description
- [Other key page](https://yoursite.com/other-page): short description
## Optional
- [Secondary page](https://yoursite.com/secondary): short description
Once the file is written, three checks are needed. First, place it at the exact root of the site (/llms.txt), never in a subfolder unless there's a specific need. Second, use full, absolute URLs for each link, not relative paths. Finally, open the address in a browser after publishing to confirm it's publicly accessible, without a password or technical block.
Mistakes to avoid
- Listing too many pages. It's not an exhaustive sitemap: keep only the pages that genuinely matter (in-depth guides, FAQs, key product pages).
- Inventing fields that don't exist. Tags like
site_name:or access rules aren't part of the specification — they're ignored or misinterpreted. - Letting the file go stale. Content that changes without an llms.txt update becomes misleading for any AI that relies on it.
- Duplicating robots.txt's role. To block a page from being crawled, edit robots.txt, not llms.txt.
- Believing the file's mere presence changes anything. Without quality content behind the links, the file adds no value.
Industry cases
A service SME's showcase site: a simple llms.txt listing the homepage, the offer, and a handful of reference articles is more than enough — the time spent on it should stay minimal given the uncertainty around its real usefulness.
Technical documentation or SaaS sites: this is the strongest use case. Documentation pages benefit more from a structured summary, especially for development agents actively navigating the site to complete a task.
Blogs or content-rich media sites: favor a selection of pillar articles rather than an llms-full.txt that tries to concatenate everything — the value of an llms.txt lies precisely in its selectivity. For this type of site, the real lever for citability remains the in-depth work covered in our guide how to get cited by ChatGPT.
Key takeaways
- llms.txt remains a proposal, not a standard adopted by the major AI players.
- It controls no access: that role belongs to robots.txt.
- It has no proven effect on Google SEO, which officially ignores it.
- Its real adoption by AI systems is close to zero today, backed by data.
- Creating it costs little: a reasonable hygiene gesture, never a visibility strategy on its own.
In summary
llms.txt is a small Markdown file that promises to guide AI on your website, but neither Google nor the leading conversational assistants rely on it yet today. Creating it takes ten minutes and carries no risk, as long as you don't confuse it with robots.txt or see it as a shortcut to AI visibility. What actually matters for being cited by a generative AI remains the quality, structure, and clarity of the content itself — that's the whole point of GEO. If you want an honest assessment of your company's real visibility in AI systems, tell us about your project: we reply within 24 business hours.
Frequently Asked Questions (FAQ)
Does llms.txt replace the XML sitemap or the robots.txt file?
No. Sitemap.xml lists every page of a site for search engines, and robots.txt handles crawling permissions. Llms.txt does neither: it offers a curated selection of content for AI systems, with no blocking or allowing function at all.
Does llms.txt improve my Google ranking?
No, according to Google Search Central's official documentation: creating this file neither helps nor hurts a site's ranking, since Google Search ignores it. It isn't a traditional ranking factor.
Do AI systems like ChatGPT actually read this file today?
Very rarely. An Ahrefs study of over 137,000 domains shows that 97% of published llms.txt files received no requests at all in May 2026, and that AI bots account for only a minority of the traffic observed on the few files that were accessed.
What's the difference between llms.txt and llms-full.txt?
Llms.txt offers a restricted selection of links to the most important pages. Llms-full.txt, a bulkier variant used by some tools, concatenates the entire Markdown content of the site into a single file — mainly useful for dense technical documentation.
Do I need a separate llms.txt file for each language of my site?
The specification doesn't require it, but if your site offers genuinely distinct content per language, one llms.txt file per language version remains more coherent than a single file mixing links across languages.
Can a poorly made llms.txt harm my site?
Not directly, since Google Search ignores it and the file has no blocking function. The risk is more subtle: outdated or misleading content in the file can mislead the few agents that do consult it.
Sources
- John Mueller (Google), reported by Barry Schwartz — Google: No AI System Currently Uses LLMs.txt (Search Engine Roundtable, June 18, 2025) — quote on the lack of use by AI systems.
- Ahrefs — We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read (June 15, 2026) — study on real-world adoption and consultation of llms.txt files.
- Google Search Central — Optimizing your website for generative AI features on Google Search — Google's official position on llms.txt.
- llms.txt specification — llmstxt.org — origin, format, and evolution of the proposed standard.
// Got a project in mind?
Let's talk about your needs.
Written by
Co-founder & Senior Advisor in Strategic Command
An entrepreneur for more than three decades, John Rademakers has helped create, grow and lead companies across a wide range of industries — from construction to aeronautics, and from automotive, finance and services to technology.
His conviction is simple: the companies that succeed over the long term rest on two inseparable fundamentals — rigorous management and effective marketing.
At NEXARA, he sets the strategic vision and guides business leaders through their decisions on digital transformation, automation and growth. Though not a developer himself, he has a deep understanding of technological challenges and relies on a team of top-level experts to design concrete, profitable solutions suited to real-world conditions.
Through his publications, he shares more than 30 years of entrepreneurial experience to help decision-makers make the right choices, avoid pointless investments and durably accelerate their growth.