There Is Already a File at Your Domain Root: Can I Customize What Appears in llms.txt?
Open a new tab, type your domain, add /llms.txt, and press enter. On Shopify you will land on agents.md. On Mintlify you will get an alphabetical list of every page in your repository. On a WordPress install running an SEO plugin you will get your five most recently updated posts. Something is already being served, and a crawler assembled it from your page tree without asking anyone in your company. So: can I customize what appears in llms.txt? Yes, on every platform in this article, and the override is usually a single file. The harder part is deciding what belongs in it once the pen is back in your hands.
Key takeaways:
- Mintlify hosts an llms.txt automatically and lists pages alphabetically in the order they appear in your repository; each description comes from the page frontmatter and truncates at 300 characters.
- Shopify stopped serving /llms.txt as an independent path on 20 May 2026 and redirects it to /agents.md; the Shopify developer changelog of 28 May 2026 introduced templates/agents.md.liquid, templates/llms.txt.liquid and templates/llms-full.txt.liquid as override points.
- Yoast SEO publishes the 5 most recently updated posts from the last 12 months plus the 5 categories holding the most content, rebuilt weekly by a scheduled action.
- Ahrefs, reading server logs from 137,000 domains, found that 97% of llms.txt files received zero requests in May 2026.
- On Mintlify, deleting your custom file restores the automatically generated one, so taking the pen back costs one commit and giving it back costs one deletion.
Go and Look at Your Domain Root Before You Read Any Further
Almost every article on this subject opens with how to author your first file. That is the wrong opening for most readers, because most readers already have one. Checking costs nine seconds and it decides everything that follows.
The scale of unauthored publication is measurable. Rankability recorded 8.7% of the world's top 1,000 websites publishing an llms.txt file in June 2026. Originality.ai, which tracks more than 3 million websites, counted 4,088 instances in June 2025 and 36,120 by May 2026, an 8.8x rise across twelve months. Curves that steep are not the sum of editorial decisions. They are platform releases reaching installed bases.
llms.txt is a plain Markdown file served at a domain root that tells an answer engine which pages carry the substance of a site, in a form it can read without rendering anything. Jeremy Howard published the proposal on 3 September 2024, and it is explicit about the difference from a sitemap: while sitemaps list all pages for search engines, llms.txt offers a curated overview for LLMs. The word doing the work there is curated. A build step that reads your page tree cannot curate; it can only mirror. Which means the real question, can I customize what appears in llms.txt, has a much shorter answer than the reading of your own file does. The sibling piece on default behaviour platform by platform maps who generates what.
Five Ways a File Appears at Your Root Without Anyone Deciding
No single vendor is responsible. Five distinct layers now ship generation as a default, and one site can be served by more than one of them.
- Documentation platforms. Mintlify hosts an llms.txt at the root of every project automatically, listing all available pages, and publishes a companion llms-full.txt carrying the full text of the documentation in a single file. GitBook, Fern and ReadMe ship comparable behaviour on publish.
- Storefront platforms. Shopify generates a discovery file for every storefront. Since 20 May 2026 the primary path is agents.md, and requests to /llms.txt redirect to it.
- CMS SEO plugins. Yoast SEO writes the file into the site root and refreshes it weekly. Rank Math adds a configurable generator. The standalone Website LLMs.txt plugin reports more than 40,000 active installations at version 8.5.3, tested up to WordPress 6.9.5.
- Edge and CDN layers. Cloudflare announced AI Index on 26 September 2025, a private beta that builds an AI-optimised index per customer domain and emits llms.txt and llms-full.txt beside a Model Context Protocol server and a search API.
- Framework modules. Static site generators read content collections directly. docusaurus-plugin-llms, vitepress-plugin-llmstxt and the Nuxt Content module emit the file at build time, so it lands in your deploy without a line in your CMS.
The shared property matters more than the list. In all five cases the input is your page tree and the output is a faithful reflection of it. Nobody chose the ordering and nobody wrote the sentences, which is why the first thing a marketing owner asks after looking is: can I customize what appears in llms.txt?
What a Generator Optimises For, and What It Cannot Know
A generator optimises for completeness and freshness. Those are the right targets for a documentation set, where the reader is an engineer hunting a specific method and any page might be the one. Mintlify's ordering rule follows from that goal: pages are listed alphabetically in the order they appear in the repository, starting from the root directory, with each description taken from the description field in the page frontmatter and truncated at 300 characters. Frontmatter is the block of key and value metadata at the top of a Markdown file. That is a clean mechanical contract, and for documentation it produces a genuinely useful file.
On a company site the same contract produces a category error. Alphabetical ordering puts About above Services. Frontmatter descriptions written for a meta tag read as marketing residue to a model looking for facts. Pages that exist for legal or navigational reasons, privacy policies, terms, tag archives, arrive with the same weight as the case study that closes deals. Yoast SEO is honest about its own axis: it includes the 5 latest updated posts and pages published in the last 12 months, priority to cornerstone content, plus the 5 categories or tags with the most content pieces attached. Recency and volume are legitimate proxies. Neither is commercial importance, and neither is what a buyer asks an answer engine.
None of this is a bug. It is the boundary of what a build step can infer. A crawler knows what exists and when it changed. It does not know which two pages you would send a serious prospect, which claim you can defend with numbers, or which product you stopped selling in March. So can I customize what appears in llms.txt is really a question about editorial ownership, not about configuration.
Reading Your Default File Like an Editor, Not a Developer
Open the file and read it as a stranger with thirty seconds and no interest in your navigation. Seven checks, in order:
- The first paragraph. Does the opening blockquote say what the business does and for whom, or does it repeat your homepage tagline? The specification asks for a short summary carrying the key information needed to understand the rest of the file.
- The top five links. Are they the five pages that carry commercial weight, or alphabetical accidents starting with About?
- The descriptions. Were they written, or lifted from a meta description field and cut mid-sentence at a character limit?
- The omissions. Is anything absent that a buyer needs, such as pricing, a case study with figures, or an integration list?
- The intrusions. Count the legal, tag and pagination URLs. Each one spends context budget on a page that answers nothing.
- The length. Mintlify caps automatically generated files at 100,000 characters and appends a note listing the number of omitted pages beyond it. If a company site approaches that ceiling, the file is a sitemap wearing a Markdown hat.
- The truth. Does it list products you no longer sell, or pages you unpublished? A stale index is worse than none, because it is confidently wrong.
Score your own file honestly. Most company sites fail checks 1, 2, 3 and 5 on a first read, and the failures share one cause: nobody wrote it. Fail two or more and can I customize what appears in llms.txt stops being a curiosity and becomes a work item.
Can I Customize What Appears in llms.txt? The Override Path, Platform by Platform
Taking the pen back is cheap everywhere in the table below. The mechanism is almost always the same: a file or template you place yourself wins over the generated output, and removing it hands generation back.
| Platform | Default output | Selection or ordering rule | Override point |
|---|---|---|---|
| Mintlify | llms.txt and llms-full.txt | Alphabetical, repository order, descriptions from frontmatter | Custom llms.txt at project root |
| Shopify | agents.md, with /llms.txt redirecting to it | Storefront routes and catalogue data | templates/agents.md.liquid, templates/llms.txt.liquid |
| Yoast SEO | llms.txt in site root, weekly rebuild | 5 newest posts and pages from 12 months, 5 largest taxonomies | Manual page selection, wpseo_llmstxt_link_description filter |
| Rank Math | Empty until configured | Up to 100 items per post type, noindex pages excluded | Post type selection plus an Additional Content field |
| Website LLMs.txt | URLs and titles only | Rebuild on content update, or daily or weekly schedule | Plugin settings and the Generate Now control |
| Cloudflare AI Index | llms.txt, llms-full.txt, MCP server, search API | Rebuilt in real time from new and updated pages | Include and exclude controls in the dashboard |
Can I customize what appears in llms.txt on Mintlify?
Yes, and the rule is stated plainly in Mintlify's llms.txt documentation: create an llms.txt or llms-full.txt file at the root of your project, and the custom file overrides the automatically generated file of the same name. Authentication changes what ships. On a fully authenticated site both files require authentication, and AI tools that cannot authenticate cannot read them. On a partially authenticated site both stay public but list public pages only.
Can I customize what appears in llms.txt on Shopify?
Yes, since 28 May 2026. Shopify's position moved twice in three weeks. Merchants reported stripped content on 8 May 2026, and on 20 May 2026 Liam, posting for Shopify on the developer community forum, confirmed that the redirect from /llms.txt to /agents.md is expected and that agents.md is now the primary storefront discovery file for agentic and AI experiences. agents.md is a sibling convention aimed at agents that act rather than only read. Eight days later the Shopify developer changelog restored control: create templates/agents.md.liquid, templates/llms.txt.liquid or templates/llms-full.txt.liquid in your theme, and if no template exists for a given path it falls back to your agents.md template, then to the Shopify-generated default.
Can I customize what appears in llms.txt on WordPress?
Yes, three ways, with three philosophies. Yoast SEO generates by rule and lets you switch to Manual page selection to override the default logic, with the wpseo_llmstxt_link_description filter for per post type descriptions. Rank Math starts from nothing: the file is empty until you configure it, then lists up to 100 items per post type, excludes anything set to noindex, and accepts arbitrary text in an Additional Content field. Website LLMs.txt sits between them, shipping URLs and titles only by default, with metadata, excerpts and taxonomies deliberately unchecked for cleaner output.
Can I customize what appears in llms.txt on Cloudflare and static frameworks?
Yes, but the control panel moves. On Docusaurus, VitePress or Nuxt the generator is a build plugin, so the override is a static file served ahead of it or a configuration key that replaces the generated title and body. On Cloudflare AI Index, the announcement commits to control over what content to include or exclude from your website's index and who can access it, managed in the dashboard rather than in your repository.
Can I Customize What Appears in llms.txt Without Losing Automatic Updates?
This is the objection that stops most teams, and it is usually misplaced. On Mintlify the trade is explicit and reversible: your custom file wins, and deleting it brings the automatically generated file back. On Shopify the fallback chain avoids an all or nothing choice, because a template covers one path while every uncovered path still falls through to the generated default. Yoast and Rank Math keep their scheduled rebuilds running while honouring your manual selections and your Additional Content block.
What genuinely does not update itself is the writing. A curated file lists perhaps fifteen to thirty URLs with one written sentence each, and those sentences go stale when the pages behind them change. Answering can I customize what appears in llms.txt with yes means giving the file an owner and a review date instead of treating it as a build artefact. Teams running several domains hit that first, which is the subject of when three platforms each generate their own.
What a Curated File Changes, Line by Line
Both files below open with the same H1 carrying the company name. Everything under it differs. First, ten lines of a generated default for a fictional freight company, Halcyon Freight:
Halcyon Freight
## Pages
- https://halcyonfreight.com/about : About
- https://halcyonfreight.com/blog : Blog
- https://halcyonfreight.com/careers : Careers
- https://halcyonfreight.com/contact : Contact
- https://halcyonfreight.com/ : Home
- https://halcyonfreight.com/privacy : Privacy Policy
- https://halcyonfreight.com/services : Services
Then ten lines of a curated file for the same company:
Halcyon Freight
> Cross border road freight between Spain and Germany for perishable goods. Founded 2009, 140 temperature controlled trucks, customs clearance handled in house.
## What we do
- https://halcyonfreight.com/services/reefer : Reefer trailers from 25 to 33 pallets, minus 25 to plus 25 degrees Celsius, Almeria to Hamburg in 38 hours.
- https://halcyonfreight.com/services/customs : In house EORI filing and phytosanitary certificates for fruit and vegetable loads.
## Proof
- https://halcyonfreight.com/cases/almeria-hamburg : 6,200 pallets a year, 99.2% on time arrival in 2025.
## Optional
- https://halcyonfreight.com/careers : Open driver and dispatch roles.
Four things changed. A positioning blockquote replaced a bare heading, so a model that reads nothing else still knows the country pair, the cargo type and the fleet size. Ordering moved from alphabetical to commercial, so the first link is revenue rather than an About page. Every link gained one written sentence carrying a number a buyer would ask for. And pages were left out on purpose, with Careers demoted into the Optional section, which the specification defines as URLs that can be skipped when a shorter context is needed. That is the practical content of can I customize what appears in llms.txt: a blockquote, an order, a sentence per link, and a set of deliberate omissions. La Boétie's own file at laboetie.io/llms.txt follows exactly that order.
The One Case Where the Default Is Already Right
Keep the generated file when your page tree genuinely is the answer. A documentation set qualifies. The reader arrives with a specific method, error code or configuration key in mind, every page is a legitimate destination, and completeness beats selection. Alphabetical ordering costs nothing when there is no commercial hierarchy to express, and frontmatter descriptions were written to describe a page rather than to sell it.
The test is one question: would a curated selection of your pages serve a reader better than all of them? For a documentation set, the honest answer to can I customize what appears in llms.txt is yes, and you should not bother. For a company site, an agency, a store or a SaaS marketing site it is almost always worth the afternoon. Where both coexist, keep automatic generation on the docs subdomain and write the file by hand at the root of the commercial domain.
How La Boétie Turns a Generated File Into a Written One
La Boétie ships the llms.txt Generator as a self serve tool at $25 or 20 € a run, with no account and no call. It exists because reading a page tree is the easy half, and writing the file is the half that needs judgement.
Read what you serve today. The run starts from your live root, records what your platform currently publishes, and reports the ordering rule and description source behind it. Most teams have never run that diagnostic.
Rewrite it as an editor. Pages are researched rather than crawled, ranked by commercial value rather than alphabet, and each link gets one written sentence with a figure in it. The difference between crawled and researched output is examined in crawled versus researched, same domain, two files.
Ship it and keep it reversible. You receive a file you own outright, ready to commit at your root or paste into a Liquid template, and deleting it restores your platform default. Ownership is the principle the studio is built on, and the same one behind Cortex, its growth engine, which produced 2 million Google impressions in three months and more than 1,000 AI citations a day for france-epargne.fr.
FAQ: Can I Customize What Appears in llms.txt on Any Platform?
Can I customize what appears in llms.txt when my platform generates it automatically?
Yes, on every platform covered here. Mintlify replaces its generated file with any llms.txt you place at your project root. Shopify reads templates/llms.txt.liquid and templates/agents.md.liquid from your theme, per the developer changelog of 28 May 2026. Yoast SEO offers Manual page selection, and Rank Math accepts arbitrary text in an Additional Content field. The mechanism differs by platform; the outcome does not.
Does adding a custom file stop the platform from updating it?
It stops automatic regeneration of that specific file, which is the point. Mintlify states that a custom file overrides the generated file of the same name, and that deleting the custom file restores the generated one. Shopify keeps generating any path you have not templated. Yoast and Rank Math continue their scheduled rebuilds around the selections and custom text you supply.
Why does my Shopify store redirect /llms.txt to /agents.md?
Because Shopify made agents.md the primary storefront discovery file on 20 May 2026, without a changelog notice at the time, which is why several developers reported stripped content on 8 May 2026 before the behaviour was confirmed. The redirect is expected. To control what is served, create templates/agents.md.liquid in your theme, and add templates/llms.txt.liquid if you want that path served separately.
How many pages should a curated file list?
Fewer than your sitemap, more than your navigation. Fifteen to thirty URLs covers most company sites: what you do, proof with figures, pricing if it is public, and the two or three explainers that answer real buyer questions. The specification provides an Optional section precisely so that lower priority URLs can be skipped when a shorter context is needed.
Is the file worth curating if almost nobody requests it?
Ahrefs found 97% of llms.txt files received zero requests across 137,000 domains in May 2026, so the file is cheap insurance rather than a traffic channel today. Curating one costs an afternoon. Serving a stale, alphabetical, legally padded index to whichever agent does arrive costs you a wrong answer about your own business.
Also worth reading: what a file worth serving contains.
Conclusion
The file at your root was written by a build step that knew your directory structure and nothing else. It is accurate and it is useless, in the way an index is accurate and useless when what you needed was an argument. Nine seconds of checking tells you which of the two you are currently serving.
So, can I customize what appears in llms.txt? Yes. On Mintlify, Shopify, Yoast SEO, Rank Math and Cloudflare the override is one file, one template or one setting, and every one of them is reversible. Do it in the order this article used: read your default critically, decide what a buyer needs, write one sentence per link, and leave things out on purpose.
Sources
Every figure and rule quoted above comes from one of the following.
- llms.txt documentation : Mintlify
- llms.txt and agents.md : Shopify Developer Community, 2026
- Customize /llms.txt, /llms-full.txt and /agents.md : Shopify developer changelog, 2026
- An AI index for all our customers : Cloudflare, 2025
- Website LLMs.txt : WordPress.org plugin directory
- Yoast SEO llms.txt functional specification : Yoast developer portal
- How to use llms.txt in Rank Math SEO : Rank Math
- The /llms.txt file : Jeremy Howard, 2024
- llms.txt adoption rises 8.8x but 97% of files get zero AI requests : PPC Land, 2026
- llms.txt adoption study : Rankability, 2026
- La Boétie llms.txt : La Boétie
Questions
Can I customize what appears in llms.txt when my platform generates it automatically?
Yes, on every platform covered here. Mintlify replaces its generated file with any llms.txt you place at your project root. Shopify reads templates/llms.txt.liquid and templates/agents.md.liquid from your theme, per the developer changelog of 28 May 2026. Yoast SEO offers Manual page selection, and Rank Math accepts arbitrary text in an Additional Content field. The mechanism differs by platform; the outcome does not.
Does adding a custom file stop the platform from updating it?
It stops automatic regeneration of that specific file, which is the point. Mintlify states that a custom file overrides the generated file of the same name, and that deleting the custom file restores the generated one. Shopify keeps generating any path you have not templated. Yoast and Rank Math continue their scheduled rebuilds around the selections and custom text you supply.
Why does my Shopify store redirect /llms.txt to /agents.md?
Because Shopify made agents.md the primary storefront discovery file on 20 May 2026, without a changelog notice at the time, which is why several developers reported stripped content on 8 May 2026 before the behaviour was confirmed. The redirect is expected. To control what is served, create templates/agents.md.liquid in your theme, and add templates/llms.txt.liquid if you want that path served separately.
How many pages should a curated file list?
Fewer than your sitemap, more than your navigation. Fifteen to thirty URLs covers most company sites: what you do, proof with figures, pricing if it is public, and the two or three explainers that answer real buyer questions. The specification provides an Optional section precisely so that lower priority URLs can be skipped when a shorter context is needed.
Is the file worth curating if almost nobody requests it?
Ahrefs found 97% of llms.txt files received zero requests across 137,000 domains in May 2026, so the file is cheap insurance rather than a traffic channel today. Curating one costs an afternoon. Serving a stale, alphabetical, legally padded index to whichever agent does arrive costs you a wrong answer about your own business.