llms.txt is a plain markdown file at the root of your site that tells an AI assistant what you do and points it at your most important pages. It takes about an hour to write. It is not robots.txt, it will not make you rank, and no major AI provider has committed to reading it. Write one anyway, because the exercise of writing it is worth more than the file.

Here is the honest position, which you will not get from anyone selling AEO as a package.

What it is

A file at yoursite.com/llms.txt, written in markdown, following the llms.txt convention. Structure is simple: an H1 with your name, a blockquote summarising what you do, then linked sections for your key pages with a line of context on each.

The idea is that a language model reading your site hits an enormous amount of navigation, cookie banners and boilerplate before it finds anything useful. llms.txt is the shortcut: here is who we are, here are the twelve pages that matter, here is what each one covers.

What it is not

It is not robots.txt. robots.txt controls access. llms.txt provides context. If you want to control which AI crawlers can read your site, that is still robots.txt, and both OpenAI and Google publish their crawler names for exactly that purpose.

It is not a ranking factor. Nobody has demonstrated that having one changes whether you get cited. Anyone telling you otherwise is guessing with confidence.

It is not widely committed to. Adoption is real and growing, but no major provider has promised to read it. That is the honest state of it in August 2026.

So why write one?

Three reasons, and only one of them is about the file.

It is cheap and the downside is zero. An hour, one file, no risk. If adoption grows you are already there.

Writing it forces a decision. To write llms.txt you must state in one blockquote what your business does, then pick the twelve pages that matter most. Most marketing teams cannot do either quickly, and discovering that is the actual value. If you cannot describe your business in three lines, an AI assistant certainly cannot.

It surfaces what your site actually says. Which brings us to ours.

Ours, and why it is wrong

We have one, live at nbh.co/llms.txt. Go and read it. Its first line describes us as a HubSpot Diamond Partner agency helping businesses grow with HubSpot CRM, marketing automation and websites.

Every word of that is true. It is also not what we are anymore.

We build AI and the revenue systems it runs on, for Australian teams of 50 to 500 people. Our own llms.txt is telling every assistant that reads it a version of us that is two years out of date, which means when somebody asks an AI which Australian firms build production AI, we are not in the answer. We are in the HubSpot answer, which we already own.

That is the useful lesson here and it applies to almost everyone reading this. The file is not the problem. The file is the diagnostic. It made us look at our own positioning in three lines and notice they were stale.

Ours is being rewritten this month. We will publish the before and after.

How to write yours

One hour, four steps.

  1. One blockquote. What you do, who for, where. If it takes more than three lines you have not decided yet.
  2. Pick the pages. Ten to fifteen. Services, the two or three pieces you actually want cited, your proof, your contact page. Not everything. Picking is the job.
  3. Annotate every link. One line saying what is on that page. This is what the file is for, and the part people skip.
  4. Put it at the root and check it loads. yoursite.com/llms.txt, plain text, no login.

Then diarise a review every six months, because a file describing a business you no longer are is worse than no file at all. Ask us how we know.

The one thing worth doing today

Open your own site's llms.txt if you have one. If you do not, write the blockquote. Just the blockquote, three lines, what you do and who for.

If it is hard, that is not a file problem, and it will not be fixed by any amount of technical AEO work.

Never named when someone asks an AI who does what you do?

Usually that is a positioning problem wearing a technical costume. We will tell you which one you have.

We are Neighbourhood. We build the AI and the revenue system it runs on. AI and RevOps engineering for Australian teams of 50 to 500 people. Diamond HubSpot Partner, Anthropic partner, 17 HubSpot Impact Awards.

Give us a shout and tell us what's broken.