Does Your Business Need an llms.txt File? A UK Guide (2026)

Type your company's domain into a browser, add `/llms.txt` to the end, and press enter. Almost every UK business gets a blank page and a 404. That missing file is a plain-text map you can hand to AI models — a curated list of your most important pages, written so a machine can read your business in seconds instead of crawling a hundred cluttered web pages to work out what you do.

Last updated: June 2026 · Covers the llms.txt proposal and how it differs from robots.txt and sitemap.xml · Written for UK SMEs, verified against Ampliflow's own live implementation

• An llms.txt file is a curated, human-readable map of your best content for AI models — not a sitemap, not robots.txt, and not a ranking trick.
• It exists because models read context windows, not whole websites — a clean map saves them from parsing your navigation, cookie banners and JavaScript to find the one thing they need.
• The honest part most vendors bury: no major AI provider has confirmed it reads the file. Its SEO or AEO payoff is unproven, full stop.
• That is precisely why it is worth doing — it costs an afternoon, it makes your business legible to machines, and it loses you nothing if the standard never lands.
• Treat it as hygiene, not magic. If anyone sells you "llms.txt optimisation" as a citation guarantee, walk away.

An llms.txt file is a curated map of your best content, not another sitemap

An llms.txt file is a single plain-text file, placed at the root of your domain (yourbusiness.co.uk/llms.txt), that tells an AI model what matters on your site and where to find it. It is written in Markdown: a heading with your company name, a one-line summary, then sections of links — each with a short description of what the page covers and why it is worth reading.

The idea came from Jeremy Howard, co-founder of Answer.AI and fast.ai, who proposed the format in 2024 and published the specification at llmstxt.org. His argument was simple. A web page is built for a browser — full of menus, banners, scripts and styling that a language model has to wade through to reach the actual content. An llms.txt file strips all of that away and hands the model the signal directly.

Think of it as the curated reading list you would give a sharp new hire on their first day. Not every document you have ever produced — the ten or twenty things they actually need to understand the business.

That curation is the whole point. A sitemap lists every URL you own. An llms.txt file lists the ones you would stake your reputation on.

It exists because models read context windows, not whole websites

To understand why the file emerged, you have to understand the constraint it solves. A language model does not browse your site the way a person does. When it answers a question, it works inside a fixed context window — a finite budget of text it can hold in memory at once.

Feed it your raw HTML and most of that budget is spent on cookie notices, navigation and tracking scripts before it reaches a single useful sentence. Feed it a clean Markdown map and every token earns its place.

This is the same instinct behind good LLM SEO: make your content easy for a model to find, lift and attribute. llms.txt takes that instinct and turns it into a file. Instead of hoping a model parses your page correctly, you hand it a pre-digested summary of what you offer and where the detail lives.

There is a fuller variant too. Some sites also publish llms-full.txt — the same map, but with the complete content of each page inlined, so a model can ingest everything in one fetch. The short file is the index; the full file is the whole library.

What llms.txt does — and the part nobody can promise yet

Here is what an llms.txt file genuinely does, today. It gives any model or agent that chooses to read it a clean, accurate, token-efficient picture of your business. It removes ambiguity about what you do, where you are and what your best content is. And it costs almost nothing to maintain.

Here is what it does not do, and the part most vendors will not lead with.

No major AI provider has confirmed that it reads llms.txt. Not OpenAI, not Anthropic, not Google. Adoption on the publishing side is real and growing — documentation platforms now generate the file automatically for thousands of sites, and major technical companies, including Anthropic's own developer docs, publish one. But publishing is not the same as consumption. Whether the models actually fetch and use these files at answer time is, as of mid-2026, unproven.

Google's John Mueller, a Search Advocate at the company, has been openly sceptical in public, noting that he has seen no evidence AI systems use the file. My own view is blunter: an undeclared, unverified signal sits uncomfortably close to the keywords meta tag — the tag search engines quietly stopped trusting two decades ago because anyone could write anything in it.

An llms.txt file is a bet that costs you an afternoon and pays off only if the industry adopts it. Tiny cost, optional upside — that asymmetry is the entire case for building one, and the entire reason to ignore anyone who sells it as a ranking trick.

State the uncertainty as a fact, not a hedge, and the decision gets easy. You are not buying citations. You are doing cheap, sensible hygiene that makes you legible to machines, on the chance that legibility starts to matter. If it does, you are ready. If it never does, you have lost an afternoon and gained a tidy map of your own business.

llms.txt, robots.txt and sitemap.xml solve three different problems

The fastest way to see what llms.txt is is to put it next to the two files it gets confused with. They share a home — the root of your domain — and nothing else. Different audiences, different jobs, different formats.

robots.txt is the older sibling: Martijn Koster proposed it in 1994 to tell crawlers which paths they may and may not visit. It governs permission. sitemap.xml is the index: a machine-readable list of every URL you want search engines to discover. It governs completeness. llms.txt is the briefing: a curated, described shortlist of what matters most. It governs comprehension.

The mistake to avoid: treating llms.txt as a second sitemap and dumping every URL into it. That destroys the one thing it is good for. A sitemap is exhaustive on purpose. An llms.txt file is selective on purpose.

How to write a good llms.txt in an afternoon

You do not need a developer or a tool. You need judgement about what matters.

1. List your 10–30 essential URLs. Your core services, your two or three best guides, your pricing, your about page, your contact route. The pages you would send a serious prospect, not the long tail.
2. Open a plain-text file. Start with a level-one heading — # Your Company — then a single blockquote line summarising what you do and where you are.
3. Group the links under `##` section headings. Sensible groups: Services, Key Guides, Key Facts, Contact. Each link is one line: - [Clear label](https://full-url): one honest sentence on what it covers.
4. Write the descriptions like a person, for a person. Specific and truthful. "24/7 AI receptionist that answers calls and books jobs" beats "innovative solutions for your needs." This is a curation exercise, not a keyword dump.
5. Save it as `llms.txt` and serve it at your root, with the content type text/plain. That is the whole technical requirement.
6. Keep it in sync. If your content changes often, generate the file dynamically so it never drifts from reality — which is exactly what we do.

What a working llms.txt looks like in production

Ours is worth opening: ampliflow.ai/llms.txt. It is not a static file someone wrote once and forgot. It is generated on the fly by a small route handler that pulls the live article list straight from our CMS and regenerates at most once an hour, so the map is never stale.

It opens with who we are and where we are — 126 High Street, Solihull — then walks a model through the business in the order a model would want it. Core services, each with a one-line description. Key facts: stack, partnerships, pricing entry points. A "priority answer targets" section that states, plainly, the questions we want to be the answer to. Hubs for our definitive Claude Code and Hermes Agent guides. Our original research. Then every article in the library, each with its excerpt, so a model can see the full depth without crawling a single page.

It also includes a section the format does not require but every business should copy: AI Crawler Access. It names the crawlers we welcome — GPTBot, OAI-SearchBot, ClaudeBot, Google-Extended, PerplexityBot and the rest — and the bulk scrapers we block in robots.txt because they take content without attribution. The map and the access rules tell one consistent story.

That is the standard to hold yourself to. Not a dumped sitemap. A deliberate, current, honest briefing on your business that any model would be glad to read — whether or not it has started reading yet.

So does your UK business actually need one?

"Need" is the wrong word. Nothing breaks without it. The right question is whether the trade is good — and for most UK SMEs it is: an afternoon of work, no ongoing cost, no downside, and real upside if the standard takes hold.

But order your priorities honestly. An llms.txt file is polish, not foundation. If the AI crawlers cannot reach your site, if your pages bury the answer three scrolls down, if you publish no data of your own, the file fixes none of that. Get the LLM SEO fundamentals right first, then add the map — with clear eyes, and without letting anyone charge premium money for something you can write before lunch.

Frequently asked questions

Is llms.txt the same as robots.txt?

No. robots.txt controls access — which crawlers may visit which paths. llms.txt aids comprehension — it hands models a curated map of your best content. They live in the same place (your domain root) and do completely different jobs. You want both.

Do ChatGPT, Claude or Gemini actually read llms.txt?

There is no confirmation that they do. No major AI provider has publicly stated that its crawlers or models consume the file at answer time. Treat any claim that a specific model "uses your llms.txt to rank you" as unproven. The honest framing is future-facing hygiene, not a live ranking input.

Where exactly do I put the file?

At the root of your domain, served as plain text: yourbusiness.co.uk/llms.txt. That is the convention, mirroring robots.txt. No sub-folders, no special headers beyond a text/plain content type.

What is the difference between llms.txt and llms-full.txt?

llms.txt is the index — headings and described links to your key pages. llms-full.txt is the same map with the full text of those pages inlined, so a model can ingest everything in a single fetch. Publish the short one first; add the full one only if your content genuinely warrants it.

Will an llms.txt file improve my SEO or AI rankings?

There is no evidence it does, and you should be wary of anyone who promises it will. Your rankings and AI citations are driven by crawlable, fast, clearly written pages with original data and clean structured data. llms.txt is a low-cost complement to that work, not a substitute for it.

How often should I update it?

Whenever your important content changes — new services, new flagship articles, a moved page. The cleaner fix is to generate it dynamically from your CMS, as we do, so it regenerates on a schedule and never drifts out of date.

My business is small. Is it worth the effort?

Yes, on the same logic that makes any cheap, no-downside bet worth taking. It is a one-off afternoon for a sole trader or a small team. Do the crawler access and answer-first basics first; an llms.txt file is a quick, sensible thing to add once those are in place.

Related reading

• ↑ From SEO to AEO: How UK Businesses Can Dominate AI-Powered Search — the pillar guide and the wider framework
• ↔ LLM SEO: Optimise for ChatGPT, Claude & Gemini — the fundamentals llms.txt sits on top of
• ↔ What Is Answer Engine Optimisation (AEO)? — the foundational definition
• ↔ Generative Engine Optimisation (GEO): The Complete UK Guide — the research and the deep tactics
• ↔ What AI Search Actually Cites for UK Business Queries (2026 Data) — our original 45-query benchmark
• ↔ The Technical SEO Audit Checklist for UK Businesses — where robots.txt, sitemaps and crawlability fit
• ↔ The AEO Checklist: 12 Moves to Get Cited by AI — llms.txt is one item; here is the rest

Start by checking what is already there

Open your browser, add /llms.txt to your domain, and look. A 404 is an afternoon's work that costs nothing and might one day matter. A thoughtful, current map of your best thinking puts you ahead of almost every competitor in Britain — quietly, which is the only way that lasts.

Get a free AI-visibility audit → — we check whether the AI crawlers can reach you, whether your pages are built to be cited, and whether an llms.txt file is worth your time before you spend a minute on it. Forty-five minutes, no commitment.

Drawing the map is the easy part. Deciding what earns a place on it is the real work — and that clarity serves you in every channel, whether a model ever reads the file or not. Draw it anyway.