What is llms.txt, and should your business bother?
llms.txt is a one-page text file that hands AI systems a curated map of your website. No engine has promised to read it yet, and it takes about an hour to write. Here is what it does, what it does not do, and when to skip it.
What llms.txt actually is
llms.txt is a single text file that sits at the root of your website, at yoursite.com/llms.txt, the same place robots.txt lives. Inside is a short, human-readable map of your site written in markdown: your business name as a heading, a one-paragraph summary of who you are, and then a handful of sections listing your most important pages with a one-line description of each.
That is the whole thing. No code, no plugin, no database. A text file you could write on your phone.
The idea comes from a proposal published at llmstxt.org. The pitch is simple: AI systems that read the web have limited room to work with, and a site full of menus, popups, and boilerplate is slow to digest. A curated index tells them where the substance lives so they do not have to guess. Think of it as the laminated menu taped up by the front door of a restaurant. Nobody is required to read it, but the ones who do walk in already knowing what you serve.
Be equally clear on the word "voluntary." Publishing an llms.txt file does not obligate any AI system to fetch it, and as of mid-2026 none of the major engines has publicly committed to doing so. More on that below, because it is the part most write-ups gloss over.
llms.txt vs llms-full.txt
The proposal describes two files, and the names confuse people, so here is the difference in one breath each.
- llms.txt is the index. A short file with links to your important pages and a sentence about each. A system reads it, understands the shape of your site, and follows the links it cares about.
- llms-full.txt is the everything file. Instead of linking to your pages, it pastes the full text of those pages into one large document, so a system can read your whole site in a single request without crawling anything.
For a local service business, llms.txt is the one worth doing first. It is small, easy to maintain, and does the core job of pointing at what matters. llms-full.txt is a reasonable add-on if your site is stable and you have a way to regenerate it when pages change. A stale everything-file is worse than none, because it misrepresents your business at length. We publish both on our own site.
What llms.txt is not, as of mid-2026
Here is the honest status, and it has not changed in the way some marketing suggests.
No major AI engine has committed to honoring llms.txt. Not OpenAI, not Perplexity, not Google, not Microsoft. The file is a proposal that some sites have adopted, not a standard the engines have agreed to. When AI assistants find and cite local businesses today, they do it through other pipelines: ChatGPT search leans on Bing's index, Perplexity fetches pages live and shows numbered citations, Google's AI Overviews pull from the regular Google index, and Gemini grounds its answers through Google Search. None of those pipelines requires an llms.txt file, and having one will not move you up in any of them.
So why bother at all? Because the file costs almost nothing, cannot hurt you, and gives any system that does choose to look a clean, curated answer to "what is this business and which pages matter." That is a real, if modest, benefit. We call it cheap insurance and a tidy front door. What we will not call it is a ranking lever, and if someone is selling it to you as one, that tells you something about the seller. For how assistants actually pick which local businesses to recommend, read our guide on how AI chooses businesses, and for the whole playbook there is the complete AI search optimization guide.
What goes in one for a local service business
The proposal came out of the software world, where llms.txt files mostly point at product documentation. A Naples plumber or a Marco Island dock builder is carrying different cargo. For a local service business, the file should answer four questions fast:
- Identity. Who you are: your about page, your contact page, anything that establishes you as a real, licensed business with a real address. This matters more than owners expect, because AI engines favor businesses they can identify unambiguously.
- Services. Your core service pages, one line each. "Water heater replacement, tank and tankless" beats "our solutions."
- Service area. The pages that say where you actually work: Naples, Bonita Springs, Estero, wherever your trucks go. Assistants field "near me" questions constantly, and geography is half the answer.
- Proof. Reviews, case studies, project galleries, licensing. The pages that let a cautious system quote something concrete about why you are trustworthy.
Ten to twenty links is plenty. This is a highlight reel, not a sitemap. You already have a sitemap.
How to write one, step by step
Honest time estimate: under an hour, most of it spent deciding which pages make the cut.
- Step 1: pick the pages. Choose ten to twenty URLs covering identity, services, service area, and proof. Skip the privacy policy, tag archives, and anything thin.
- Step 2: write the summary. Two or three sentences: who you are, what you do, where you work, and one credibility fact like your license or founding year.
- Step 3: format it in markdown. Your business name after a # symbol, the summary after a > symbol, section headings after ##, then one line per page: a hyphen, the page name in square brackets, the full URL in parentheses, a colon, and a short description.
- Step 4: save and upload. Save it as llms.txt, plain text, not a Word file, and upload it to your site's root folder so it loads at yoursite.com/llms.txt. Whoever hosts your site can do this in minutes.
- Step 5: check it loads. Type the URL into a browser. You should see raw, unstyled text. If you get a 404 or a styled page, it is in the wrong place.
- Step 6: keep it current. When you add or retire an important page, update the file. A stale map sends readers to dead ends.
A worked example
Here is a finished file, line by line, for a fictional company we will call Gulf Breeze Plumbing. It is not a real business; we invented it for this example. Each bullet below is one line of the actual text file:
- # Gulf Breeze Plumbing
- > Licensed, family-run plumber serving Naples and Bonita Springs, Florida since 2009. Repipes, water heaters, leak detection, and 24-hour emergency service. Florida license CFC000000.
- ## Services
- - [Water Heater Replacement](https://example.com/water-heaters/): tank and tankless installs, same-week scheduling
- - [Whole-Home Repipes](https://example.com/repipes/): copper and PEX repipes for older Naples homes
- - [Leak Detection](https://example.com/leak-detection/): non-invasive slab and wall leak location
- ## Service Area
- - [Naples Plumber](https://example.com/naples/): our home base, serving all of Collier County
- - [Bonita Springs Plumber](https://example.com/bonita-springs/): south Lee County coverage
- ## About
- - [About Us](https://example.com/about/): the family behind the company, licensing and insurance
- - [Reviews](https://example.com/reviews/): what customers say about our work
That is the entire file. The # and > and ## symbols are markdown, a plain-text formatting convention AI systems parse easily. Swap in your own pages, keep the descriptions honest and specific, and you are done.
How it complements robots.txt
These two files get confused because they live in the same place, but they do opposite jobs.
robots.txt is the bouncer. It tells crawlers who may come in and which rooms are off limits. llms.txt is the greeter. For the crawlers you have already let in, it says: here is what to look at first. One controls access, the other offers orientation, and neither replaces the other. llms.txt has no power to grant or deny anything.
That ordering matters in practice. If your robots.txt or your firewall blocks AI crawlers, your llms.txt is a welcome mat behind a locked door. Nothing ever reads it. Before spending even the one hour, spend five minutes confirming AI systems can reach your site at all. Our walkthrough on checking whether AI can read your site shows exactly how to test, and if the test fails, the unblocking guide covers the fix. It is common for a security plugin or CDN setting to switch on AI-bot blocking by default without the owner ever noticing.
Where to see a real one
Ours is live. Type naplesseo.com/llms.txt into any browser and you will get the raw text file, no styling, no page around it: our name and summary at the top, then sections for services, guides, locations, and results, each a list of links with a one-line description. We also publish naplesseo.com/llms-full.txt, the expanded version, and our robots.txt explicitly allows the AI crawlers that matter, including GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, and Google-Extended. We practice what this page describes, which also means you are welcome to copy our structure.
When llms.txt is a waste of time
We just told you it is cheap. Cheap is not the same as always worth it. Skip it, or at least postpone it, in these situations:
- Your site blocks AI crawlers. Fix access first. A curated map behind a locked door helps nobody, and the fix usually matters far more than the map.
- Your site has almost nothing to map. If you have a five-page brochure site, the honest priority is building real service pages, service-area pages, and proof pages. Write the llms.txt after there is something worth pointing at.
- It is being sold to you as a ranking play. Any proposal that puts llms.txt ahead of your Google Business Profile, your reviews, and answer-shaped service pages has the priorities backwards. Walk away.
- It is substituting for structured data. Schema markup is a standard that Google actually consumes today. llms.txt is a proposal nobody has committed to. If you only do one, do schema for AI search.
- It is delaying real work. One hour is fine. A week of fiddling while your service pages sit thin is not.
The bottom line for Naples business owners
If your fundamentals are in order, write one. It takes about an hour, it cannot hurt you, and it leaves a clean front door for whichever AI systems come knocking next. Just keep it in its lane: a small good habit, not a strategy. Nothing in AI visibility moves overnight, and we set a 90-day ramp expectation with every client; llms.txt is one hour inside that ramp, not a shortcut around it.
If you would rather not touch a text file, our done-for-you llms.txt and crawlability service handles the authoring plus the part that matters more, an audit of whether AI crawlers can reach your site at all. And if you just want a quick read on where you stand right now, the free AI Visibility Checker takes about a minute.
Frequently asked questions
See where you stand in Google and AI search
We will run a free audit of your local rankings and your visibility across ChatGPT, Gemini, Perplexity, and Google AI Overviews, then show you the fastest path to more booked jobs.