Everyone's Adding llms.txt to WordPress. Does It Actually Do Anything?
What llms.txt is, what it is not, who honours it, and how to publish one that says exactly what you want AI systems to know.
In this article 8 sections
llms.txt is a plain-text file, served at the root of your site (yoursite.com/llms.txt), that gives AI systems a curated, human-written map of your most important content: what your site is, what it covers, and which pages best represent it. It is a proposed convention, not an accepted web standard — no major AI provider has committed to honoring it — but it costs almost nothing to publish, and any system that does read it gets your best content described in your own words rather than inferred from crawling.
This guide covers what llms.txt actually is, its honest status in 2026, what to put in one, a realistic example, how it fits alongside robots.txt and schema, and how to publish one from WordPress in a few minutes.
What llms.txt is (and where it came from)
The proposal is simple: large language models and the retrieval systems around them consume web content differently from search crawlers. They work best with concise, structured, plain-text context — and a full HTML site, with its navigation, footers, scripts and boilerplate, is a noisy way to deliver that context.
llms.txt addresses this by offering a single Markdown-formatted file at a well-known location. The proposed format is:
- An H1 with the site or project name
- A short blockquote summarizing what the site is
- Optional free-form notes
- One or more H2 sections containing lists of links to key pages, each with a brief description
Think of it as a hand-written sitemap for machines that read prose: not everything you have, but the pages you would show a new reader who asked, "what should I read to understand this site?"
The honest status: proposed, promising, unproven
Clarity matters here, because plenty of coverage overstates the case.
What is true: llms.txt is a public proposal with a defined format. A meaningful number of documentation sites and software companies have adopted it. Publishing one is harmless — it is a static text file that changes nothing about how your site renders or ranks.
What is not established: No major AI provider — OpenAI, Anthropic, Google, Perplexity — has publicly committed to fetching and honoring llms.txt as part of their standard pipelines. Unlike robots.txt, which has decades of established (and now formalized) crawler behavior behind it, llms.txt has no enforcement mechanism and no guarantee of consumption.
The rational posture: treat it as a cheap, asymmetric bet. Ten minutes of work, zero downside, and a real upside if the convention gains adoption — plus an immediate side benefit: writing one forces you to articulate what your most important pages are, which is a useful exercise regardless. It slots naturally into a broader answer engine optimization effort as the cheapest single item on the list.
What to include in your llms.txt
A good llms.txt is curated, not exhaustive. Guidelines that hold up in practice:
- Name and describe the site in one or two sentences. Write it for a reader with zero context. This summary may be the only thing a system quotes about you, so make it accurate.
- Link your cornerstone content. The pages that best answer the questions your audience asks: core guides, documentation, product pages, your about page. Ten to thirty links is a sensible range for most sites.
- Describe each link briefly. One line saying what the page covers and who it is for. These descriptions are your chance to frame each page's purpose explicitly.
- Group links into sections. Docs, guides, product, company — whatever taxonomy fits. An "Optional" section is conventionally used for secondary material a system can skip when context is tight.
- Skip the noise. Tag archives, pagination, thin posts, legal boilerplate. If you would not hand it to a new reader, do not hand it to a model.
- Keep it current. A stale llms.txt pointing at renamed or deleted pages is worse than none. Revisit it when your cornerstone content changes.
A realistic example
Here is what a complete llms.txt might look like for a small WordPress business site — a fictional pottery studio:
# Kiln & Clay Studio
> Kiln & Clay is a ceramics studio in Portland offering wheel-throwing
> classes, open studio memberships, and handmade tableware. This file
> lists our most useful pages for AI systems answering questions about
> our classes, pricing, and policies.
## Classes
- [Beginner Wheel Throwing](https://kilnandclay.example/classes/beginner):
6-week course for complete beginners; schedule, pricing, what to bring.
- [Intermediate Techniques](https://kilnandclay.example/classes/intermediate):
Trimming, handles, and glazing for students with wheel experience.
- [Class FAQ](https://kilnandclay.example/classes/faq): Cancellation
policy, age limits, gift certificates, and makeup sessions.
## Studio
- [Open Studio Membership](https://kilnandclay.example/membership):
Monthly membership tiers, included firing, and studio hours.
- [Firing Services](https://kilnandclay.example/firing): Bisque and
glaze firing for outside work, pricing per shelf.
## About
- [About the Studio](https://kilnandclay.example/about): Our story,
instructors, and location with parking details.
- [Contact](https://kilnandclay.example/contact): Booking questions
and studio visits.
## Optional
- [Blog](https://kilnandclay.example/blog): Studio news and student
showcases.
Notice what makes this work: a system reading only this file could answer most real questions about the business correctly, or at least know exactly which page to fetch next. That is the entire job.
How llms.txt fits with robots.txt and schema
These three mechanisms are complementary, not competing — each answers a different question a machine might ask:
| Mechanism | Question it answers | Status |
|---|---|---|
| robots.txt | "Where am I allowed to go?" | Long-established, formalized |
| XML sitemap | "What exists on this site?" | Long-established |
| Schema (JSON-LD) | "What is this specific page?" | Established, documented by Google |
| llms.txt | "What matters here, and how should I understand it?" | Proposed convention |
robots.txt is access control: it tells crawlers (including AI crawlers, which increasingly identify themselves with distinct user agents) what they may fetch. It says nothing about what is worth fetching. Manage it carefully — the robots.txt editor guide covers the WordPress side, and robots meta settings handle per-page directives.
XML sitemaps enumerate everything; llms.txt curates. A sitemap with 2,000 URLs is useful to an indexer and useless as context. llms.txt is the opposite trade.
Schema markup describes individual pages with machine-readable precision — authorship, dates, FAQ structure, product data. It operates page by page, where llms.txt operates site-wide. If you are choosing where to invest first, schema has the stronger evidence behind it; the schema markup guide makes that case in full.
A site that wants to be legible to AI systems publishes all four, and they take perhaps an afternoon combined.
Publishing llms.txt from WordPress with Hydrogen SEO
You can create the file by hand and upload it to your web root, but on WordPress that means filesystem access and remembering the file exists every time content changes. The Hydrogen SEO plugin includes a dedicated llms.txt editor that handles it from wp-admin:
- Install the free plugin (installation guide) — the llms.txt editor works without an account, like every non-AI feature.
- Open the llms.txt editor from the Hydrogen SEO dashboard.
- Start from the provided default content and edit it into your curated map — name, summary, sections, links.
- Save. The file is served at yoursite.com/llms.txt with no file transfers involved.
- Verify by visiting the URL directly in a browser.
The llms.txt editor documentation walks through each screen. Developers can also filter the output programmatically via the plugin's hydrogen_seo_llms_txt_content hook — see the filters reference.
The same plugin covers the neighboring mechanisms in the table above — the robots.txt editor with validation warnings, XML sitemaps, and no-code schema markup — so the whole "legible to machines" layer lives in one place, server-rendered with zero front-end JavaScript.
Common questions
Does llms.txt affect Google rankings? No. It is not a ranking signal, and nobody credible claims otherwise. Its purpose is context for AI systems, not search ranking.
Can llms.txt block AI training? No. It is informational, not access control. If you want to restrict crawlers, that is a robots.txt job, with the honest caveat that robots.txt is a request, not an enforcement mechanism.
Should I list every page? No — curation is the point. Machines already have your sitemap for exhaustiveness.
Is there a companion format? The proposal also describes offering Markdown versions of individual pages and a fuller llms-full.txt variant. These remain even less adopted than the core file; start with llms.txt itself.
The bottom line
llms.txt is a ten-minute task with an honest risk profile: certainly harmless, possibly valuable, and clarifying to write either way. Publish one, keep it current, and spend the time you saved on the fundamentals with proven weight — structure, schema, and content worth citing.
If you are on WordPress, Hydrogen SEO gives you the llms.txt editor, robots.txt editor, sitemaps and schema in one free plugin. Create a free account if you also want the AI tools and visibility monitoring on top — the file editor itself needs nothing but the plugin.