llms.txt SEO for Small Business Websites: Should You Add One?

If you run a small business website, you’ve probably heard someone mention llms.txt SEO lately. It’s a small text file that sits at your website’s root, and it’s supposed to help AI systems understand your site. Some people call it essential. Others call it a waste of time. The honest answer is somewhere in between — and it depends entirely on what kind of site you run.

This guide explains what llms.txt actually is, which AI systems read it (and which ones openly ignore it), what a good one looks like, and how to decide whether your business should bother. We’ll also show you our own live implementation, because we don’t recommend anything we haven’t done ourselves.

What llms.txt Is, in Plain English

The llms.txt proposal came from Jeremy Howard, co-founder of Answer.AI, who published it on September 3, 2024 at llmstxt.org. The idea was deliberately simple: give AI models a clean, curated map of your website so they don’t have to guess what’s important.

Think about your site from an AI’s point of view. When a tool like ChatGPT or Perplexity needs to answer a question about your business, a crawler visits your site and sees raw HTML — navigation menus, footers, cookie banners, popups, sidebar widgets. The actual content that matters is buried in that noise. Your XML sitemap doesn’t fix this: it lists every URL equally — the homepage, the privacy policy, that landing page from 2021 you forgot to delete. There’s no hierarchy, no context, no “start here.”

llms.txt fixes that by handing AI systems a short Markdown document that says, in effect: here’s what this site is about, and here are the pages worth reading. That’s the whole concept. A curated table of contents for machines that read.

A quick reality check on its status: llms.txt is a community proposal, not a W3C or IETF standard. It’s still maintained at llmstxt.org as open for community input. Compare that with the two files you already know:

The fence analogy is the one that sticks: robots.txt is the fence around your property, sitemap.xml is the phone book listing every room, and llms.txt is the welcome brochure in the lobby.

Which AI Systems Actually Read It

This is the section most guides gloss over, so let’s be straight about it.

Google Search ignores llms.txt. Google’s own guide to generative AI features, last updated July 10, 2026, says so plainly: the file neither helps nor harms your visibility in Google, including AI Overviews and AI Mode. OpenAI, Anthropic, and Perplexity document their crawlers without mentioning it. A study by SE Ranking across nearly 300,000 domains found no measurable effect on AI citations from having an llms.txt file. If someone tells you this file will get you cited by ChatGPT, ask them for evidence — the public evidence points the other way.

So who does read it? Adoption is real, just in specific corners of the web. Anthropic, Vercel, Cloudflare, Stripe, Supabase, and Zapier all publish llms.txt files. Mintlify rolled it out across thousands of documentation sites in late 2024, which is where most of the current adoption lives: developer tools and SaaS documentation. More than 780 websites now ship one, and community directories track the growing list. AI assistants, coding agents, and developer tools that fetch documentation on demand are the primary readers.

Then there’s the Lighthouse angle, which deserves its own section because it’s the one development that changed the conversation in 2026.

The Lighthouse Angle: Google Audits It, But Doesn’t Rank It

In May 2026, Google added an llms.txt check to Chrome Lighthouse’s new “Agentic Browsing” audit category. That made llms.txt an official readiness check — a real audit you can run in DevTools today.

The Agentic Browsing category evaluates how well your site is built for machine interaction. It checks four things: the accessibility tree (the same structure screen readers use), layout stability (your CLS score), any WebMCP tools your site exposes, and the presence of an llms.txt file. Instead of a 0-to-100 score, it reports a pass ratio — like 2/3 or 3/3.

Here’s the part people miss: the category carries no ranking weight. Google has been explicit that it exists to gather data while agentic-web standards are still forming. A perfect score doesn’t make you rank higher, and it doesn’t get you cited by AI systems either.

So why did Google audit something its own search team says to skip? SEO consultant Lily Ray put the question directly to Google’s John Mueller, noting that Google itself publishes llms.txt files. Mueller’s answer drew a clean line: “discovery” (getting found by a search engine) versus “functionality” (helping an agent or a user do something once they’ve arrived). Search indexing and agentic browsing are two different futures. In the first, an AI answers from a pre-built index and your llms.txt never enters the pipeline. In the second, an autonomous agent lands on your actual site to complete a task, and a machine-readable map of your content saves it time. llms.txt lost the first race before it started. It may yet matter in the second.

One technical gotcha worth knowing: if you do publish a file, Lighthouse’s parser is strict about markdown. The audit has failed sites with the message “File does not appear to contain any links” even when the file listed pages in plain text — because the links weren’t written in markdown syntax with square brackets and parentheses. A missing file is simply marked “not applicable.” A malformed one can actually fail. If you’re going to do it, do it right.

What a Good llms.txt Contains

The format is short on purpose. The proposal defines a simple structure:

1. An H1 with your site name — the only strictly required part.
2. A blockquote summary — one or two sentences on what the site covers and who it’s for.
3. H2 sections of annotated links — each line written as - [Page title](URL): one-line description. Group them by topic: services, guides, pricing, about, contact.
4. An optional “Optional” section — pages tools can skip if they want, like your changelog or blog archive.

There’s also a companion convention called llms-full.txt: the full text of your key pages in a single Markdown document, so a tool can load everything at once. It’s optional, not part of the core proposal.

Curation is the entire point. You pick the 10 to 20 pages that represent your site, rather than letting a crawler infer importance from your link structure. That means: your main service pages, your pricing, your about page, your best guides — and nothing else. Leave out the privacy policy, the terms page, the tag archives, and anything you’d never hand to a prospect.

We run our own implementation at seoelinks.com/llms.txt. It opens with our name and a one-paragraph summary of the agency, then groups links under clear headings: services (technical SEO, ecommerce, content strategy, link building, audits), how we work (our process, pricing, FAQ), and contact. Every link carries a one-line description written the way we’d describe the page to a person, because that’s exactly what a language model reads. It takes about fifteen minutes to write a file like this, and ours is deliberately short — a file nobody maintains is worse than no file at all.

Who Benefits — and Who Can Skip It

You probably benefit if: your site has documentation or an API that developer tools and AI assistants reference constantly; you run a store or SaaS product with dozens or hundreds of pages, where a curated brief helps assistants find pricing, policies, and comparisons without crawling everything; or you want a clean, quotable summary of your business available to any AI tool that bothers to look. For these sites, llms.txt is a fifteen-minute investment with genuine forward-looking value.

You can skip it if: your site has five pages and a contact form. A good sitemap and well-written pages already give AI systems everything they need — a five-page site doesn’t need a map. You can also skip it if you’re expecting a ranking boost or a flood of AI citations. That is not what this file does, and anyone selling it as an SEO lever is selling you something the data doesn’t support.

The middle ground — a typical small business site with 15 to 50 pages — is a judgment call. It costs nothing to serve and a quarter of an hour to write. Keep it current, and it’s a reasonable bet on the agentic web. Let it go stale, and it’s just another file to maintain for no return.

Your llms.txt Implementation Checklist

If you’ve decided it’s worth doing, here’s the full checklist:

1. Draft the content in Markdown. H1 with your business name, a blockquote summary of what you do and who you serve, then H2 sections grouping your 10–20 most important pages as annotated links.
2. Write descriptions like a human. Each link gets one sentence describing what the reader will find. Avoid keyword stuffing — this is a map, not a ranking tactic.
3. Curate ruthlessly. If a page doesn’t help a stranger understand or buy from you, leave it out. Fifteen great links beat fifty mediocre ones.
4. Use real markdown link syntax. Every link must be [Title](https://yoursite.com/page/) — square brackets, parentheses, full URL. Lighthouse fails files that skip this.
5. Save it as llms.txt and upload it to your domain root — it must load at yoursite.com/llms.txt. That’s the only location the convention recognizes.
6. Verify it loads. Open yoursite.com/llms.txt in a browser. You should see clean Markdown, not a 404 page or your homepage.
7. Check it in Lighthouse. Run Chrome DevTools’ Lighthouse with the Agentic Browsing category ticked. The llms.txt audit should read “follows recommendations.” If it fails on links, fix the syntax from step 4.
8. Keep it current. Revisit the file when you add or remove major pages. A stale map sends AI tools to dead ends — set a calendar reminder twice a year.
9. Don’t block AI crawlers in robots.txt while expecting AI visibility. That’s the most common self-inflicted wound in this space: people lock the door, then wonder why nobody visits.

The Bottom Line

llms.txt SEO is a small, honest bet. It won’t move your Google rankings — Google says so itself. It probably won’t get you cited by ChatGPT tomorrow — the studies say so too. But it costs fifteen minutes, it makes your site legible to the AI agents and tools that are already visiting it, and Google’s own Lighthouse now checks for it as part of the agentic web’s emerging infrastructure.

For most small business websites, that’s enough reason to add one — as long as you do it properly and keep it fresh. The file is the welcome brochure, not the building. Your content, your reviews, and your actual expertise still do the heavy lifting.

Want a second opinion on your site’s AI readiness? We’re SEOelinks — an SEO agency built in Brampton, Canada, serving clients worldwide. Take a look at our own llms.txt at seoelinks.com/llms.txt, then request a free audit and we’ll tell you straight what your site needs next.

Written by

SEOelinks Team

SEOelinks is a Brampton-based SEO agency helping service businesses across Canada and worldwide win organic search. Our strategists publish the same field-tested playbooks on technical SEO, local visibility and AI search that we run for clients every day.

More about SEOelinks →