SerpTune
EN

← Blog

llms.txt: what it is, how to create it, and what the evidence says

llms.txt is a plain Markdown file served at the root of your domain — like robots.txt, but addressed to AI assistants instead of crawlers. It hands language models a curated map of your most important pages in a format they can read without parsing your HTML. Proposed in 2024, by 2026 it sits on most SEO checklists — usually with more enthusiasm than evidence. Here is both the how and the honest what-for.

What goes in the file

The convention is deliberately minimal: one H1 with the site name, a short blockquote describing what the site offers, then sections of annotated links.

# ExampleShop
> Direct-to-consumer store for hiking gear, with sizing guides and repair tutorials.

## Guides
- [Boot sizing guide](https://example.com/sizing/): how to measure and choose a size
- [Repair tutorials](https://example.com/repairs/): fix zips, soles and membranes

List the pages that answer questions — guides, documentation, pricing, FAQ — not every URL you own. Keep the annotations factual: the file is read by a model, not ranked by an algorithm, so adjectives buy you nothing.

How to create and serve it

  • Static or generated. A hand-written file works for small sites; anything that changes weekly should generate it from the same source of truth as your sitemap, so it never goes stale. SerpTune serves its own at serptune.com/llms.txt, rebuilt on every request.
  • Plain text, UTF-8, root path. No HTML wrapper, no redirect chains.
  • Do not block AI crawlers in robots.txt while inviting them in llms.txt — assistants that respect one tend to read both.

What the evidence says in 2026

The honest part: no major AI provider has officially confirmed consuming llms.txt, large-scale tests across hundreds of thousands of domains have found no measurable traffic change for the overwhelming majority of adopters, and adoption among top-traffic sites remains in the low single digits. Anyone selling llms.txt as a ranking lever is ahead of the data.

Why we serve one anyway

Because the cost is near zero and the failure mode is harmless. If assistants start honoring the convention, early files are already in place; until then, it documents your site for the growing share of visits that begin inside a chat window. Treat it as a low-cost hedge — not a strategy.

What actually moves AI citations today

The measurable levers remain on the page itself: an extractable answer at the top of each section, clean structure, verifiable numbers and named sources — the 8 citation signals we score. Run any URL through the free AI Overview checker first; write the Markdown map after.

Sign up