If you have spent any time reading about AI search, you have probably bumped into the term llms.txt and wondered whether it is the next big thing or just another file you are supposed to feel guilty about not having. Here is the honest version: llms.txt is a simple, useful idea that takes about an afternoon to set up, and it belongs near the bottom of your priority list, not the top.
In plain English, llms.txt is a Markdown-formatted text file that lives at the root of your website, the same place robots.txt and sitemap.xml live. Its job is to hand large language models a short, human-curated guide to your best content so they do not have to guess which of your pages actually matter. Think of it as a concierge for AI: instead of letting a model wander your entire site, you point it straight to the rooms worth visiting.
What is llms.txt, exactly?
The llms.txt standard was proposed in 2024 by Jeremy Howard as a way to make websites easier for AI systems to read. Web pages are noisy. They are full of navigation menus, cookie banners, scripts, ads, and boilerplate that mean nothing to a buyer but cost a model time and context to parse. An llms.txt file strips all of that away and presents a clean outline in Markdown, a lightweight format that models handle especially well.
A basic llms.txt file has a predictable shape:
- An H1 with the name of your business or site.
- A short blockquote summarizing who you are and what you do.
- One or more sections (H2 headings) grouping related pages.
- Under each section, a Markdown list of links to your key pages, each with a one-line description.
That is the whole thing. There is a companion idea called llms-full.txt that bundles the actual text of those pages into one file, but for most businesses the lean version is plenty. The point is curation, not volume. You are telling AI, “Of everything we publish, these are the pages that explain us best.”
llms.txt vs robots.txt vs sitemap.xml
People mix these up constantly, so it helps to line them up side by side. They look similar because they all sit at your domain root, but they do very different jobs.
| File | Audience | What it does |
|---|---|---|
| robots.txt | Crawlers | Tells bots what they may not access |
| sitemap.xml | Search engines | Lists every URL you want indexed |
| llms.txt | AI models | Highlights your best pages in clean Markdown |
The key contrast: robots.txt restricts, sitemap.xml inventories, and llms.txt recommends. A sitemap is a flat list of everything; llms.txt is an opinionated short list of what matters most, written so a model can absorb it quickly. They are complementary, and you can run all three on the same site without conflict.
Does ChatGPT, Claude, or Google actually read llms.txt?
This is the question that matters, and the honest answer is: not in any officially confirmed way yet. As of early 2026, none of the major answer engines has publicly stated that llms.txt is a retrieval or ranking input. It is a community-driven proposal that has gained traction among developers and documentation sites, but it is not an adopted standard the way robots.txt is.
That does not make it worthless. Adoption of new web standards almost always runs ahead of official confirmation, and the cost of adding the file is trivial. But it does mean you should treat llms.txt the way we do across the audits we run: as a small clarity signal, not a channel you can count on. If anyone promises that an llms.txt file will get your business cited by ChatGPT, be skeptical. The systems that decide who gets recommended are far more interested in your reviews, your structured data, and whether your content directly answers the questions people ask.
If you want the deeper mechanics of why that is, our explainer on how AI assistants decide who to recommend walks through the signals that genuinely drive AI recommendations today.
Does your business actually need an llms.txt file?
For the vast majority of small and local businesses, llms.txt is optional and low priority. It will not hurt you, and it is a tidy thing to have, but it is not where your attention should go first. Here is a rough way to think about who benefits most:
- Documentation-heavy and SaaS sites get the clearest value. If you have hundreds of technical pages, an llms.txt file genuinely helps models find the canonical ones.
- Large content libraries benefit because curation cuts through the noise. If you publish a lot, pointing AI to your anchor pages is worthwhile.
- Small service businesses get the least lift. With ten or twenty pages, a model can already see your whole site easily. The file is a nice-to-have, not a need-to-have.
If you run a mortgage practice, a law firm, an accounting office, or a real estate team, the file takes an afternoon and you can add it as a finishing touch. Just do not let it crowd out the work that moves the needle. The fundamentals of answer engine optimization matter far more: clean schema markup, a complete and consistent Google Business Profile, real customer reviews, citations across the right directories, and content that answers buyer questions directly and early.
How to create an llms.txt file
If you have decided it is worth doing, the process is short. You do not need a developer for the basic version.
- Create a plain-text file named exactly
llms.txt(lowercase). - Add an H1 with your business name, for example
# Riverside Mortgage Advisors. - Add a one-paragraph blockquote summary describing what you do and who you serve.
- Group your best pages under H2 sections such as Services, Locations, and Guides.
- List each page as a Markdown link with a short description, like
- [First-Time Buyer Loans](https://example.com/first-time-buyers): How we help new buyers qualify and close. - Upload it to your domain root so it loads at
yoursite.com/llms.txt. - Keep it current. When you launch important pages or retire old ones, update the file so it never points AI at dead links.
Keep it focused. A tight file of your fifteen best pages does more good than an exhaustive dump of everything you have ever published. The goal is to make your most useful content unmistakable.
A quick note on what llms.txt cannot do
An llms.txt file cannot manufacture authority you have not earned. It cannot conjure reviews, fix inconsistent business listings, or make thin pages worth citing. We saw this firsthand with the work that took Seattle mortgage broker Keith Akada from invisible in AI results to the number-one AI-recommended broker in his market, with roughly 30 leads and four closed deals inside six weeks. That result came from the fundamentals working together, not from a single file. llms.txt was, at most, a small piece of a much larger system.
Where llms.txt fits in a real AEO strategy
The right way to frame llms.txt is as a clarity layer on top of work that already exists. If your site has strong, answer-first pages, accurate structured data, and consistent citations, then an llms.txt file gives AI a tidy front door to all of it. If those fundamentals are missing, the file is a beautifully labeled door to an empty room.
Our priority order, refined across the audits we run, looks like this: get your answer engine fundamentals right first, then add structured data, then strengthen reviews and citations, and only then add finishing touches like llms.txt. Done in that order, every step compounds. Done out of order, you are polishing the doorknob on a house with no foundation.
So should you make an llms.txt file? If it takes you an afternoon and your fundamentals are already solid, sure. Just go in with clear eyes: it is a sensible tidy-up, not a growth lever. The businesses that win in AI search are the ones that became genuinely worth recommending, and then made that easy for machines to see.