llms.txt: What It Is, Who Reads It, and Ours — Annotated

By Ivan Siyanko 7 min read

An llms.txt file is a plain-text, Markdown-formatted file placed at the root of a website (at /llms.txt) that gives large language models a curated, machine-readable summary of the site and links to its most important pages. It was proposed as a convention to help AI systems find and understand a site's key content quickly, in the same spirit as robots.txt or a sitemap, but aimed at LLMs rather than crawlers. Adoption is early and uneven: publishing one is low-cost and low-risk, but you should not assume every AI engine reads it today, and claims that it guarantees citations are unfounded. It is best treated as one useful signal within generative engine optimisation (GEO), not a magic switch. This guide explains what llms.txt is, who actually consumes it as of July 2026, whether it helps your AI visibility, how to write a good one, and it annotates NeuralGen's own live file line by line. If you want the honest position: implement it as part of a broader GEO setup, and measure — do not expect it to work in isolation.

On this page
  1. What is llms.txt?
  2. Who actually reads llms.txt?
  3. Does llms.txt help AI visibility?
  4. What does a good llms.txt look like?
  5. NeuralGen’s llms.txt — annotated
  6. How do you create and publish an llms.txt file?

What is llms.txt?

An llms.txt file is a root-level Markdown file, published at /llms.txt, that summarises a site and links to its key pages for large language models to read.

It is a proposed convention, not an official standard mandated by any single AI engine — analogous to robots.txt or a sitemap, but written for models rather than crawlers. NeuralGen is a London AI-visibility agency, and this file is one of the concrete, implementable artefacts of the discipline we practise. See our generative engine optimisation guide and LLM SEO pillar for how it fits into the wider picture.

The problem it tries to solve is real, even if the solution’s adoption is still forming. A website built for human visitors is full of navigation, marketing copy and layout that a model has to wade through to find the actual facts. An llms.txt file skips straight to a curated index: here is who we are, and here are the pages that matter, in one place, without the surrounding noise.

The name itself is a slight misnomer worth clarifying. It is not read exclusively by large language models in some technical sense — it is simply a file written with an LLM reader in mind, structured for extraction rather than for a human scrolling a webpage. Anyone, human or machine, can open it and understand the site in seconds.

Who actually reads llms.txt?

Adoption is early and uneven, and you should not assume universal support across AI engines. <!-- CONFIRM: current llms.txt consumption by ChatGPT, Perplexity, Claude and Gemini as of July 2026 — verify each before publishing; do not overstate. -->

No engine has publicly confirmed that it uses llms.txt as a ranking or citation input, and this status should be treated as directional rather than settled. That is not a reason to skip it — it costs little to publish — but it is a reason not to oversell it to a client or to yourself.

This is a common pattern with emerging web conventions: robots.txt and sitemaps both took time to become universally respected, and llms.txt is following a similar early path, proposed by practitioners before being formally adopted by every major platform. The honest position is to publish it as good practice and low-cost insurance, while putting your main effort into the signals that are better evidenced — entity consistency, quotable content and genuine corroboration. Our generative engine optimisation guide covers that fuller set of signals.

Does llms.txt help AI visibility?

It may help models find and summarise your key content, at very low cost and risk, but it is one signal among many and not a substitute for entity consistency, quotable content, schema and citations.

Treat it as a supporting fix, not a headline strategy. NeuralGen guarantees measurement and method, not a citation — that applies here as much as anywhere else. Our guide to AI visibility explains the metric you should actually be tracking to know whether any of this is working.

If you publish an llms.txt file and your AI visibility score does not move, that is not proof the file was pointless — it is more likely evidence that a different signal, such as entity inconsistency or thin content, is the actual bottleneck. Isolating the effect of any single fix is difficult, which is exactly why measuring the whole picture, rather than any one change in isolation, is the more reliable approach.

Get your free AI Visibility Scorecard

What does a good llms.txt look like?

A good llms.txt is concise, factual and cleanly structured: an H1 with the site or brand name, a short blockquote summary, then sectioned links to key pages with a one-line description each.

# Brand Name

> One-sentence, accurate description of what the company does.

## Docs
- [Page title](/url/): one-line description of what this page covers.

## Guides
- [Page title](/url/): one-line description of what this page covers.

## Key pages
- [Page title](/url/): one-line description of what this page covers.

The H1 names the entity unambiguously. The blockquote is the one sentence you want quoted back to you, so it should match your standard company description everywhere else. Each section groups links by purpose, and each description should stand alone if extracted out of context.

Keep the whole file short. The point is a curated index, not a mirror of your sitemap — listing every page on your site defeats the purpose, because it forces a model back into the same wading-through-noise problem the file was meant to solve. A dozen well-chosen links with accurate one-line descriptions is more useful than a hundred links with none.

NeuralGen’s llms.txt — annotated

Here is NeuralGen’s own live file, with the reasoning behind each part.

# NeuralGen

> NeuralGen is a London AI-visibility agency. We get UK brands cited and
> recommended by ChatGPT, Perplexity, Claude and Google AI — and prove it
> with monthly share-of-voice reports.

## Services
- [/ai-seo-agency/](/ai-seo-agency/): AI visibility and GEO service — get found and recommended by AI engines.
- [/ai-automation/](/ai-automation/): fixed-price AI automation for SMEs.
- [/partners/](/partners/): white-label GEO/AI-SEO fulfilment for agencies.
- [/academy/](/academy/): practical AI training for UK teams.

## Guides
- [/generative-engine-optimisation/](/blog/generative-engine-optimisation/): what GEO is and how it works.
- [/llm-seo/](/blog/llm-seo/): the broader discipline of optimising for large language models.

## Get started
- [/ai-visibility-scorecard/](/ai-visibility-scorecard/): free AI visibility scorecard, 48-hour turnaround.
- [/pricing/](/pricing/): every price on the site, stated openly.

<!-- CONFIRM: paste NeuralGen's actual live /llms.txt contents here before publishing; keep annotations aligned to the real file. -->

The summary uses the exact company boilerplate from the master brief, kept identical to the description on the site and elsewhere, because consistency is the trust signal a model cross-references. The links are grouped by purpose — services, guides, get started — rather than dumped as a flat list, so a model (or a person) can scan the structure in seconds. Every description is written to stand alone, in case only that line is extracted.

Two choices are worth calling out specifically. First, the summary sentence is verbatim identical to the description NeuralGen uses on its homepage and in its schema markup — this is not an accident but the whole point: a model encountering the same sentence in three different places has stronger grounds to treat it as fact. Second, the “Get started” section deliberately leads with the free scorecard rather than a sales page, because the file’s job is to help a model understand and route people usefully, not to read like an advert.

How do you create and publish an llms.txt file?

Creating one is a short, mechanical job: list your key pages, write one accurate description of your company, format it as Markdown, and publish it at your site root.

  1. List your most important pages — services, pillar guides, about, pricing. Resist the temptation to include everything; a curated list beats a comprehensive one.
  2. Write a one-line, accurate company description, and reuse it consistently everywhere — your homepage, your schema, your directory listings, and this file.
  3. Format as Markdown: H1 brand name, blockquote summary, sectioned links with descriptions, following the shape shown above.
  4. Save the file as llms.txt and upload it to your site root so it resolves at /llms.txt — most content management systems allow this with a static file upload; no special plugin is required.
  5. Keep it current when pages change, and treat it as part of your standard content checklist rather than a one-off task.
  6. Measure your AI visibility before and after with our scorecard, so you learn what actually moves rather than assuming.

If you’d rather have this built and measured for you, that is part of what our AI SEO agency service does.

A common mistake is treating this as a one-off task rather than a maintained artefact. A file that still links to a discontinued service page, or describes a company using an old strapline, actively works against you — it is one more inconsistent source for a model to weigh against your current, correct pages. Whoever owns your website content should treat updating llms.txt as part of the standard checklist whenever a key page changes.

Get your free AI Visibility Scorecard

Frequently asked questions

What is an llms.txt file?

An llms.txt file is a plain-text, Markdown file at a website's root (/llms.txt) that gives large language models a curated summary of the site and links to its most important pages. It is a proposed convention to help AI systems find and understand key content quickly — similar in spirit to robots.txt, but aimed at LLMs.

Do AI engines actually read llms.txt?

Adoption is early and uneven as of July 2026, and you should not assume every engine reads it. No engine has publicly confirmed it uses llms.txt as a ranking input. Publishing one is low-cost and low-risk, so it is worth doing as part of a broader GEO setup — but measure rather than assume it works alone.

Does llms.txt help my AI visibility?

It may help models find and summarise your key content, but it is one signal among many, not a magic switch. Entity consistency, quotable content, schema and citations in trusted sources matter more. Treat llms.txt as a low-cost addition to a broader generative engine optimisation setup, and measure the effect.

Where do I put the llms.txt file?

Save it as a Markdown file named llms.txt and upload it to your website's root so it resolves at yourdomain.com/llms.txt. Structure it with an H1 brand name, a short summary, and sectioned links to your key pages with one-line descriptions. Keep it updated when important pages change.

Is llms.txt the same as robots.txt?

No. robots.txt tells crawlers which URLs they may or may not access. llms.txt gives large language models a curated, human-readable summary and links to your most important content. They serve different purposes and can coexist; llms.txt does not control access, it offers guidance and context.

Keep reading