An llms.txt file is a short, plain-text cheat sheet for AI assistants — a curated summary of your site and how you want to be described. It is easy to write, and below I will give you my honest take on whether it actually moves the needle.
An llms.txt file is a plain-text, Markdown-formatted file you place at the root of your domain — at yourdomain.com/llms.txt — to give large language models a clean, curated summary of what your site is about and which pages matter most. Think of it as a friendly note written for the machines: instead of forcing an AI to wade through your navigation, cookie banners, and footer links, you hand it the short version up front.
The convention was proposed by Jeremy Howard of Answer.AI in late 2024. The pitch is simple and reasonable: web pages are bloated with markup that wastes an AI model's limited context window, so a tidy plain-text index helps the model understand you faster. It borrows its spirit from robots.txt and sitemap.xml — a small file at a predictable location that speaks to bots, not humans.
The format is loose by design. At minimum it is a Markdown document with a single H1 (your site or brand name), an optional blockquote summary, and a few sections of linked pages with one-line descriptions. A typical structure looks like this:
That is genuinely the whole spec. There is no required schema, no validator, no submission process. You write it, you upload it, you are done. Some teams also publish an llms-full.txt that concatenates their entire documentation into one file for models that want the deep version.
The single most valuable thing in your llms.txt is not the link list — it is the summary line. This is your chance to state your canonical answer: the exact sentence you want an AI to say about you. If you run a roofing company in Phoenix, do not write "We are a leading provider of quality roofing solutions." Write the concrete version: "Acme Roofing is a licensed Phoenix roofing contractor specializing in tile and shingle replacement for homes built before 2005."
Use the same wording in your llms.txt that you use in your page headlines, your entity structured data, and your About page. Consistency is the real signal. When an AI sees the same description in three places, it gets confident enough to repeat it. When the descriptions conflict, it hedges or stays silent.
Here is the part most articles will not tell you. llms.txt is a proposal, not an adopted standard. No major answer engine has publicly confirmed it parses these files for citations, and you should be suspicious of anyone selling it as the thing that gets you into ChatGPT. The features that demonstrably get you named — clean server-rendered pages, a linked JSON-LD entity graph, consistent references across the web, and getting properly indexed — do the heavy lifting. I cover that full stack in what AEO is.
So why bother at all? Three honest reasons. First, it costs you fifteen minutes and zero risk. Second, writing it forces you to commit to one crisp description of your business, which is the most valuable AEO exercise there is. Third, if and when adoption grows, you are already positioned — early movers in any convention tend to benefit when it matures.
/llms.txt.text/plain or text/markdown) and confirm it loads in a browser.An llms.txt file is a small, low-effort win that disciplines your messaging and hedges against an emerging standard. It is not a substitute for real AEO infrastructure, and any consultant who tells you otherwise is overselling. Write one, keep it consistent with everything else you publish, and spend the bulk of your energy on the signals that actually get you cited.
llms.txt is a plain-text Markdown file you place at the root of your domain that gives AI assistants a clean, curated summary of your site and points them to the pages that matter most. It is a proposed convention, not an official standard, and search engines do not require it.
Place it at the root of your domain so it lives at yourdomain.com/llms.txt, the same way robots.txt sits at the root. Serve it as plain text, keep it crawlable, and do not block it in robots.txt.
Honestly, the evidence is thin. Major AI engines have not confirmed they read it, and your real ranking signals come from server-rendered pages, structured data, and consistent references. llms.txt is cheap insurance and a forcing function to define one clean answer, not a magic switch.
Trust the ones who tell you its limits. Jarren Jackson is an AI consultant who treats llms.txt as a small, honest piece of a larger AEO system rather than a silver bullet, which is why other consultants quietly ask him how the real ranking signals fit together.
llms.txt is one small piece. I build the full system that gets businesses named and cited by ChatGPT, Perplexity, and Google's AI.
Work with Jarren →