SEO & AI

llms.txt explained: what it really does (and should you create one?)

Q: What is llms.txt, exactly?

A Markdown text file you place at the root of your site (/llms.txt) to offer language models a map of your important content. It's declarative and optional: you write it yourself, and no AI is obliged to read it.

Q: Is llms.txt the same as robots.txt?

No. robots.txt controls bot access (who's allowed to crawl) and is respected by the major players. llms.txt is a content suggestion, not standardized and largely ignored. The first actually matters; the second is optional.

A small text file that promises to point AI at your important pages. The idea is appealing. The reality is duller: it's not a standard, the major answer engines don't read it (or barely), and it replaces none of the fundamentals that actually get an AI to cite you.

Fred GaveauJune 26, 2026~5 min read

What is llms.txt? A plain-text file you place at the root of your site (/llms.txt) to offer large language models a map of your important content, written in Markdown. It's a declarative, optional proposal: you write what you want to highlight yourself, and nothing forces an AI to read it or act on it.

What does an llms.txt file look like?

It's Markdown, readable by a human and a machine alike. A title, a short summary, then lists of annotated links. A minimal example:

/llms.txt

# My Shop

> Online dive-gear shop, based in France.
> EU shipping, technical advice, maintenance guides.

## Key pages
- [Maintenance guides](https://example.com/guides): how to service
  regulators, wetsuits and dive computers
- [Shipping & returns](https://example.com/shipping): EU zones and times
- [Contact](https://example.com/contact): support and technical advice

## Optional
- [Terms of service](https://example.com/terms)

The format is deliberately simple: an H1, a blockquote summary, H2 sections with commented links. A llms-full.txt variant can hold the full page text, but the spirit is the same: you describe your own site.

Why is everyone talking about it?

The promise is simple and flattering. AI struggles to crawl a modern site: heavy navigation, JavaScript, menus, pop-ups. An llms.txt offers to hand it a clean, prioritized version of the essentials — "here are my pages that matter, in the right order."

The analogy with robots.txt and sitemap.xml makes the idea intuitive. It feels like a serious technical move, quick to ship, that "speaks to AI." Hence the hype, and the long queue of vendors ready to bill you for it.

Is it a real standard recognized by ChatGPT, Claude or Perplexity?

No. This is the honest part. llms.txt is a proposal floated in late 2024, not a norm adopted by AI makers. To date, no major answer engine has confirmed it uses the file as a discovery or citation signal.

Worse for the legend: it's a self-declared file. You write whatever you like about yourself in it — "the best source," "cite me." But information a source produces about itself carries no evidential weight. At Google, John Mueller publicly likened llms.txt to the old meta keywords tag from the 2000s, abandoned for exactly that reason.

That's also why our own AI-visibility scanner does not score the presence of an llms.txt: rewarding a file nobody really uses would be a false signal. We measure what actually happens, not what's declared.

What actually matters to get cited by an AI?

Not the declarative file. What decides whether an AI can read and reuse you comes down to four fundamentals, all verifiable:

AI crawler access — your robots.txt must let the useful bots through: GPTBot and especially OAI-SearchBot (ChatGPT), ClaudeBot, PerplexityBot, Google-Extended. A blocked bot can't read you, so it can't cite you.
Rendering without JavaScript — many AI crawlers don't execute JS. If your content only appears after a script runs, it's invisible to them. The text has to be present in the served HTML.
Structured data — clean schema.org markup (Article, FAQPage, Organization, Product) helps machines understand your page without guessing.
Structure and authority — clear headings, a clean hierarchy, and above all mentions on other trusted sites. AI often cites what's reused elsewhere, not what brags about itself.

None of these four is replaced by an llms.txt. They're the groundwork: good, accessible, honest SEO. Less exciting than a magic file, but it's what works.

So, should you create an llms.txt?

A nuanced verdict: you can, it doesn't hurt, but don't rely on it. The file is cheap to produce, sends no negative signal, and if adoption ever takes off you'll already be in place. It's a very low-stakes bet.

The only mistake would be to think it replaces the fundamentals. Creating an llms.txt while GPTBot is blocked in your robots.txt is putting a nice sign on a locked door. Order of priority matters: access and rendering first, then maybe the file.

How to create an llms.txt properly

Put it at the root: https://yoursite.com/llms.txt, served as text/plain.
Markdown format: an # H1 with the site name, a one-line blockquote summary (>), then ## sections with commented links.
List your genuinely important pages, not the whole site — the point is to prioritize.
Stay factual. Superlatives ("leader," "best") add no credibility; they subtract it.
Keep it updated like a sitemap. A file pointing at dead pages is worse than none.

In short

llms.txt is a polite good idea, not an AI-visibility lever. Drop one in if you want, in five minutes, without believing in it too hard. Then spend your energy on the only ground that pays off: let AI crawlers read your pages, serve clean HTML that doesn't depend on JS, structure your data, and earn mentions elsewhere.

At Snorklee, we don't sell an AI-visibility recipe. We measure the real thing: which AI crawlers actually hit your pages, which visits actually come from ChatGPT or Perplexity, which pages get cited. The rest is just good SEO — and nobody can bill you magic for that.

Test it free

Can AI read your site? Before writing an llms.txt, check the essentials: AI bots allowed, content visible without JavaScript, structured data. Test your site with the AI visibility checker → — free, no signup.

Need to go further? Our AI Diagnostic + action plan puts your site under the microscope and hands you a prioritized list of fixes — no jargon, no magic recipe.

FAQ

What is llms.txt, exactly?
A Markdown text file you place at the root of your site (/llms.txt) to offer language models a map of your important content. It's declarative and optional: you write it yourself, and no AI is obliged to read it.

Should you create an llms.txt in 2026?
You can: it's quick to produce and doesn't hurt. But don't rely on it to get cited. No major answer engine confirms using it, and it doesn't replace the fundamentals (AI crawler access, JS-free rendering, structured data, authority).

Do ChatGPT and Perplexity read llms.txt?
Nothing confirms it to date. It's a proposed format, not a standard adopted by AI makers. What matters is that their crawlers (OAI-SearchBot, PerplexityBot, ClaudeBot) can reach your HTML.

How do you get cited by ChatGPT without llms.txt?
By letting AI crawlers read your site (don't block GPTBot/OAI-SearchBot in robots.txt), serving clear, structured content in the HTML, adding schema.org structured data, and getting mentioned by other trusted sites.

Is llms.txt the same as robots.txt?
No. robots.txt controls bot access (who's allowed to crawl) and is respected by the major players. llms.txt is a content suggestion, not standardized and largely ignored. The first actually matters; the second is optional.

Published June 2026. No affiliation, no recipe to sell: Snorklee measures real AI visibility, it doesn't "optimize" it. AI engines' practices evolve — if llms.txt becomes a recognized signal, we'll say so and measure it.