If you've spent any time around web and SEO conversations lately, you've probably heard someone mention llms.txt like it's the next robots.txt, a magic file you drop on your website so ChatGPT and the other AI tools will finally understand your business and recommend you to customers.
Here's the honest version: llms.txt is a real proposal, it's cheap to add, and it might help. But it is not a standard, the major AI companies have not committed to honoring it, and anyone selling it to you as a guaranteed visibility boost is ahead of the evidence. Let's walk through what it actually is, how it relates to the files that do matter, and what a small business owner should do about it.
First, the file it gets confused with: robots.txt
Every website can have a small text file at its root called robots.txt. It has been around since the 1990s, and it's how a site tells automated visitors, called crawlers or bots, which parts of the site they're welcome to read. Googlebot checks it before crawling your pages. So do Bing's crawler, OpenAI's crawlers, and most other legitimate bots. Google documents how this works in its crawler and robots.txt documentation, and it's worth knowing the basics even if you never touch the file yourself.
Two things to understand about robots.txt:
- It's about permission, not understanding. It says "you may read this" or "please don't read this." It does not explain your business, your services, or your service area.
- It's voluntary. Well-behaved crawlers respect it. Bad actors ignore it. It's a posted sign, not a locked door.
robots.txt is settled, widely honored, and worth getting right. llms.txt is a different animal.
What llms.txt actually is
The llms.txt idea was proposed in late 2024 by people in the AI developer community. The pitch is simple: websites are messy for machines. They're full of menus, popups, scripts, and layout code. When an AI system tries to read your site, it has to dig the actual content out of all that noise.
So the proposal says: put a plain, human-readable text file at yoursite.com/llms.txt that acts as a curated table of contents for AI systems. In plain language, it would say something like "here's who we are, here's a short description, and here are links to our most important pages in a clean, easy-to-parse format." Some sites go further and publish full plain-text versions of their key pages alongside it.
Think of it as the difference between handing someone your entire filing cabinet versus handing them a one-page summary with tabs pointing to the documents that matter.
That's the whole idea. It's not code, it's not complicated, and a competent web person can produce one in under an hour.
What llms.txt is not
This is where I want to be straight with you, because there's a lot of overselling happening.
It is not a standard. robots.txt is honored by essentially every major crawler. llms.txt is a proposal that some sites have adopted and some tools can read. As of this writing, none of the major AI companies, including OpenAI, Anthropic, or Google, have publicly committed to using llms.txt as an input to their assistants or search products. Some crawlers may fetch the file; fetching is not the same as using.
It is not access control. llms.txt doesn't block anything or grant anything. If you want to tell AI crawlers to stay out, that's a robots.txt job, and we wrote a whole companion piece on whether you should: Should You Block AI Crawlers?.
It is not a ranking factor. Nobody at Google has said llms.txt affects your search results, and there's no mechanism by which it would. Google's guidance for showing up in search, including its newer AI-powered features, still comes down to crawlable pages, useful content, and structured data, all covered in their search documentation.
It is not a substitute for a good website. If your site doesn't have a page for each service you offer, clear contact information, and your service area spelled out in actual text on actual pages, a summary file pointing at thin content summarizes nothing.
So why would anyone bother?
Because the cost is nearly zero and the downside is nearly zero, and the upside, while unproven, is plausible.
Here's the plausible part. AI assistants increasingly answer questions like "who's a good HVAC company near Wilmington?" or "how much does duct cleaning usually cost?" To answer, they rely on what they can read and understand from the web, partly through their own crawling and partly through traditional search indexes like Google's and Bing's. Anything that makes your site easier for a machine to read correctly works in your favor across all of those channels.
llms.txt is one small bet in that direction. The bigger, proven bets are:
- Clean, crawlable pages. One page per service, real text, fast loading. The fundamentals haven't changed.
- Structured data. This is machine-readable markup that's actually documented and actually used by Google and Bing today. It deserves its own article, and we wrote one: Structured Data: Feeding the Answer Engines.
- Consistent business information. Same name, address, and phone number everywhere your business appears online.
If you've already done those three, adding llms.txt is a reasonable "why not" move. If you haven't, do those first. Putting a polished table of contents on a broken filing cabinet helps no one.
What goes in one, conceptually
If you or your web person decide to add the file, keep it simple and honest:
- A one-line description of the business. Plain English: what you do, where you do it.
- A short paragraph of context. Your services, your service area, anything an assistant would need to describe you accurately.
- A list of your most important pages with one-line descriptions. Your services pages, your about page, your contact page, your pricing page if you publish one.
Notice what that list looks like: it's the same information a good homepage already communicates. That's not a coincidence. Machines and humans want roughly the same things from your site, which is clarity about who you are, what you do, and where.
One caution: don't stuff it. Some people treat llms.txt as a place to dump keywords and marketing copy hoping AI systems will repeat it. There's no evidence that works, and the history of search is one long lesson that systems eventually learn to ignore or punish manipulation. Write it like you'd brief a new employee, not like you're gaming a slot machine.
How this fits with the other "AI files" you'll hear about
A quick map, because the terminology gets sloppy:
- robots.txt controls which crawlers may read your site, including AI crawlers like OpenAI's GPTBot, which OpenAI documents on its own site. Settled and respected.
- Sitemaps tell search engines which pages exist so nothing gets missed. Settled and respected.
- Structured data (schema markup) labels what your content means: this is a business, this is a service, these are our hours. Settled, documented at schema.org, and actively used by search engines.
- llms.txt offers a curated summary for AI systems. Early, unsettled, optional.
Three of those four are table stakes for a small business website in 2026. The fourth is a cheap option you take after the first three are done.
Our take for local businesses
If you run an HVAC company, a plumbing shop, a roofing crew, or any local service business, here's the priority order we'd give you over coffee:
- Make sure your site has a real page for every service you want to be found for. Our industry pages for HVAC and plumbing show the shape of this.
- Make sure search engines and AI crawlers can actually read your site. Most can, but plenty of DIY builders and old sites quietly block or break crawlers.
- Add structured data so machines know your hours, location, services, and reviews without guessing.
- Then, if you want, add llms.txt. It takes an hour, it can't hurt, and if the convention catches on you're already there.
That order matters. We've reviewed a lot of small business sites where the owner was asking about llms.txt while their services were buried in a single paragraph on the homepage. The exotic stuff is fun to talk about. The boring stuff is what gets you found.
And if the convention dies quietly? You've lost an hour and gained a clearly written summary of your business, which is useful to have anyway.
Want this handled for you?
This is exactly the kind of thing we build into every site by default. Omnyra is a veteran-owned web shop in Wilmington, NC, and we've built 1,500+ small business sites in the last 90 days using a done-with-you process: we build your site live on a call with you, you get a first draft in 24 hours, and you're live in 7 days, guaranteed.
Structured data and AI-search visibility are built into our Standard tier at $2,000 plus $200 a month for hosting, maintenance, and monthly content. Tiers run from $500 up to Super Max from $6,000, and pay-in-4 or Klarna financing is available on all of them. See the full breakdown on our pricing page, or book a call and we'll look at your current site together, no charge and no pressure.