llms.txt: what it is, and whether your UK small business actually needs one

If you spend any time reading about AI search, you will have run into the phrase llms.txt by now, usually presented as the new thing every website must add or be left behind. That framing is overheated. llms.txt is real, it is worth understanding, and for many businesses it is worth adding. It is also a proposal rather than a settled standard, and it is not the deciding factor in whether an AI assistant cites your business. This is the honest version: what it is, how it differs from the files you already have, where adoption actually stands, and whether your UK small business needs one.

This piece sits under our wider guide to AI search optimisation, which covers the whole question of getting a small business named inside Google’s AI Overviews, ChatGPT and other answer engines. If you want the foundations first, start there. If you came here specifically for llms.txt, read on.

What is llms.txt?

llms.txt is a proposed plain-text file you place at the root of your website, at yourdomain.com/llms.txt, that tells large language models what your site is about and points them to your most important content in a clean, easy-to-read form.

The idea, proposed in 2024 by Jeremy Howard of Answer.AI, is simple. AI models work from a limited context window, and a full web page is cluttered with navigation, scripts, cookie banners and markup that get in the way of the actual content. An llms.txt file gives a model a curated, distraction-free map: a short description of your business, then a tidy list of links to the pages that matter, each with a one-line note on what it covers. It is written in Markdown, so it is readable by a person and a machine alike.

A small llms.txt for a plumbing business might open with the business name and a sentence on what it does, then list the services page, the service-area page, the pricing page and a few key guides, each with a short description. Nothing more complicated than that. The file does not contain secrets or instructions; it is a friendly summary that says “here is who we are, and here is the good stuff”.

How is it different from robots.txt and sitemap.xml?

This is the question worth getting straight, because the three files sound similar and do entirely different jobs. You can have all three, and they do not conflict.

• robots.txt controls access. It is the long-established file that tells crawlers which parts of your site they may and may not visit. It is a gatekeeper, written for crawlers, and every search engine respects it. If you want to allow or block AI crawlers specifically, robots.txt is where that happens, not llms.txt.
• sitemap.xml is a complete inventory. It lists every URL on your site so search engines can discover and index them all. It is exhaustive by design, machine-only, and built for traditional search crawling.
• llms.txt is a curated summary. It does not list everything; it highlights the handful of pages that best explain your business, in a form a language model can read cheaply and cleanly. It is editorial where a sitemap is mechanical.

Put plainly: robots.txt says what a crawler is allowed to touch, sitemap.xml says where everything is, and llms.txt says what is worth reading and what it means. The first two are settled web standards. The third is a proposal trying to become one.

Is anyone actually using it? The honest state of play

Here is where credibility matters, because this topic attracts a lot of confident nonsense. The honest position as of mid-2026 is that llms.txt is a community proposal with real momentum but no official endorsement from the major AI companies.

A growing number of developer tools, documentation sites and forward-leaning businesses publish an llms.txt file, and there are public directories listing thousands of sites that have adopted it. That is genuine adoption. What does not exist, at the time of writing, is public confirmation from Google, OpenAI, Anthropic or the other large providers that their crawlers read llms.txt and weight it when assembling an answer. Some have said little; some have indicated they do not currently use it. It is a standard the community is proposing to the AI companies, not one they have agreed to follow.

So treat anyone who promises that adding llms.txt will get you cited in ChatGPT as either mistaken or selling something. Nobody outside those companies can promise that, because the mechanics are undisclosed and the file is not yet a confirmed input. What you can say honestly is that it is a sensible, low-cost bet on where things are heading, and that it does no harm while you wait to find out.

Why this matters at all: the shift to answer-first search

The reason llms.txt exists is the same reason AI search optimisation exists. The way people find businesses online is changing. Roughly 60 percent of Google searches now end without a click on any result, because the searcher gets their answer on the results page itself (SparkToro). At the same time, ChatGPT reports around 900 million weekly active users (OpenAI), many of whom now ask it the kind of question they would once have typed into Google.

In both cases a machine reads across sources and writes a single answer, naming only a few of them. Being one of those named sources is the new version of ranking on page one. llms.txt is one small attempt to make your site easier for those machines to read correctly. It is not the whole job, but it fits the direction of travel, and that is why it is worth taking seriously even while it is unproven.

Who actually benefits from an llms.txt file?

It helps some sites more than others. Be realistic about which you are.

• Documentation-heavy and content-rich sites benefit most. If you publish guides, a knowledge base, or a large library of articles, an llms.txt file that points models at the canonical version of each topic is genuinely useful, because it cuts through clutter and reduces the chance of a model quoting the wrong or outdated page.
• Businesses with a clear set of key pages benefit modestly. A typical local service business, a handful of service pages, a pricing page, an about page and a few guides, can produce a clean llms.txt in minutes that summarises the business well.
• Tiny brochure sites benefit least. If your whole site is five pages and already clear, llms.txt adds little a well-structured site does not already provide. It will not hurt, but it is not where your effort should go first.

The honest hierarchy is this: a clear, well-structured site that states its facts plainly is worth a hundred times more than any text file, because that is what the models actually read. llms.txt is a tidy bonus on top of good foundations, never a substitute for them. If your pages bury the answer in paragraph four, an llms.txt file pointing at them will not save you. Fix the pages first.

How to add one (it is genuinely an afternoon)

If you have decided it is worth doing, the process is straightforward.

1. Write a short summary of your business. One heading with your name, then a sentence or two on what you do and who you serve. Plain English, no marketing fluff.
2. List your most important pages. Pick the handful that best explain the business: services, service area, pricing, about, and your best guides. Write each as a Markdown link with a short note on what it covers.
3. Save it as llms.txt in Markdown. The structure is a top-level heading, an optional blockquote summary, then sections of links. The proposal’s own site documents the exact format if you want to be precise.
4. Upload it to the root of your domain so it lives at yourdomain.com/llms.txt, exactly where robots.txt and sitemap.xml sit.
5. Keep it current. Like any summary, a stale one is worse than none. When your services or prices change, update the file. This is the part most people forget, and it is the part that decides whether the file stays useful.

That is the whole job. There is no submission process and no approval; you publish the file and it is there for any crawler that chooses to read it.

So, does your UK small business need one?

Here is the plain answer. No, you do not need an llms.txt file, in the sense that nothing breaks without one and no AI company currently requires it. But it is a low-effort, low-risk addition that aligns with where search is heading, so if you have an afternoon spare it is a reasonable thing to add, particularly if you publish guides or have a content-rich site.

What you should not do is treat it as the headline of your AI-search effort. The work that actually moves the needle is the unglamorous foundation: stating your answers plainly near the top of each page, marking up your facts with structured data, building real credibility signals, and keeping fast, current pages. That is the substance the models read regardless of any text file, and it is covered properly in our companion pieces on schema for AI answer engines and how to get cited by ChatGPT and AI Overviews. The full nine-discipline picture, including AI Search SEO, is in the free SEO playbook, and the existing breakdown of why small businesses are invisible in Google’s AI Overviews shows the four gaps that matter most.

If all of this sounds like a lot to keep on top of, that is the honest truth of modern search: it is a moving target, and the files, the markup and the freshness all need a hand on them. That is exactly what a managed website service is for, one person who builds the site properly and keeps these foundations current so you do not have to track every new proposal yourself.

Where to start

If you want to know how your current site stands on the things that genuinely decide AI-search visibility, structure, speed, clarity and credibility, the fastest first step is a free Site Score. It tells you where you are strong and where you are leaking, in plain English, before any commitment. Or if you would rather talk it through, get in touch and we will look at your site together and tell you which of this is worth your time and which is noise.