Ai Search
What an llms.txt File Actually Does
An llms.txt file is a plain markdown file you place at the root of your domain, at /llms.txt, that hands AI models a clean, curated map of your most important pages instead of making them parse your full site. It is a proposed convention, not a rule any engine must follow, and that distinction matters far more than most vendors will tell you.
This page explains what a line in an llms.txt file actually does, where the format genuinely helps, and where it does nothing yet. We build and maintain these files as one part of an AI search engagement. We would rather be straight about their limits than sell you a file that sits ignored at the edge of your site.
What an llms.txt file actually contains
The format is deliberately simple, which is part of the point. The file opens with a single H1 that names your business or project. A blockquote beneath it gives a one or two sentence summary of what you do. Then come H2 sections, each holding a markdown list of links, and every link carries a short description of what a reader will find on the other side.
A companion file, llms-full.txt, goes further by concatenating the full text of your key pages into a single long document, so a model can read one file instead of following links. Both are written in markdown because markdown is cheap for a language model to read. It strips away the navigation, scripts, cookie banners, and markup that normally bury the few sentences that actually describe your business.
What a line in the file does, and what it does not do
Here is the part the marketing skips. Putting a URL in your llms.txt file does not force any model to read that page, cite it, or rank it. The file is a suggestion. An engine still has to choose to request /llms.txt, parse it, and decide to trust it. The major engines have not publicly committed to using the file as a retrieval or ranking signal, and at least one large search provider has said in public that it does not consume it today. Anyone promising guaranteed citations from an llms.txt file is selling certainty that does not exist.
It also helps to be precise about what you are chasing. A citation is when an AI answer links to or names your page as a source. A mention is when the answer describes you without attribution. An llms.txt file influences neither directly. What it can do is make your best content easier to retrieve and summarize accurately, which is a precondition for both. The file supports the outcome. It does not deliver it.
So why build one? Because the cost is low, the downside is near zero, and the upside is real if adoption grows. Writing the file is also a forcing function. It makes you decide which pages matter, describe each in one honest sentence, and see how your entities connect. That clarity strengthens the rest of your AI search work whether or not a single model ever opens the file.
llms.txt, robots.txt, and your sitemap are not the same
These three files get confused because they sit near the root of your domain, but they do different jobs. Your robots.txt controls access and tells crawlers, including AI crawlers, which paths they may fetch. Your sitemap.xml is an exhaustive index that lists every URL you want discovered, with no editorial judgment. An llms.txt file is the opposite of exhaustive. It is a short, curated list that says, in effect, start here, these pages explain us best.
Order matters here. An llms.txt file is pointless if your robots.txt quietly blocks the very AI crawlers you want to reach, which is a common and easy mistake to make. We handle that specific problem on /services/ai-search-optimization/ai-crawler-access/.
How we build and maintain your llms.txt file
We run AI search optimization only on sites we build, so your llms.txt file is generated from a site whose structure, headings, and entities we already control. That coupling is the whole advantage. The file is never better than the pages it points to, and a tidy link to a vague, unstructured page helps no one.
The file is never the whole job. It sits alongside content structuring, entity work, and schema, because a curated map only helps when the destinations are worth reading. This runs on a published monthly retainer with a 90 day initial term, with the custom website included, and we take one client per industry per city so your map never competes with a file we wrote for a rival. The structure is laid out on /services/seo/ and /pricing/seo/.
Common mistakes that make the file useless
Most llms.txt files fail quietly. They are generated once, never updated, and slowly drift out of step with the site. Others try to list every page, which defeats the entire purpose of curation. The worst ones describe pages in marketing language a model cannot safely repeat, so even if the file is read, nothing useful comes out of it.
Who should build one now, and who can wait
If you publish documentation, a deep catalog of services, or dense reference content, an llms.txt file is worth building now, because it gives a model a clean doorway into material that is otherwise hard to parse. If you run a small local service site with a handful of pages, the file is quick to add but will move little on its own. Either way it is a supporting move, not the headline. The real question is whether your whole site is readable by machines, and this file is one line item inside that larger answer.
We do SEO on websites we build. AI search optimization is part of that same engagement.
Questions
Do I actually need an llms txt file?
Not urgently, and not on its own. If your site has deep documentation or a large catalog, it is a cheap, sensible addition that gives models a clean entry point. If you run a small site, it will do little by itself and should come after your content and crawler access are sorted. Think of it as a supporting file, never the centerpiece of an AI search strategy.
Does ChatGPT or Google actually read llms.txt?
There is no public commitment from the major engines that they use llms.txt as a retrieval or ranking signal, and at least one large search provider has said publicly that it does not consume the file. Adoption may grow, which is why building a clean file now is a cheap hedge. But treat any promise of guaranteed citations from the file as a red flag.
What is the difference between llms.txt and llms-full.txt?
llms.txt is a curated map: an H1, a short summary, and lists of links with descriptions. llms-full.txt goes further and concatenates the full text of your key pages into one long markdown document so a model can read everything in a single pass. You can publish one or both. The map is the starting point and the full file is optional depth.
Can an llms.txt file get my business cited by AI?
Not directly. The file cannot force a citation or a mention. What it can do is make your best content easier for a model to find, read, and summarize accurately, which is a precondition for being cited at all. The citation itself depends on the quality and structure of the pages the file points to.
How often does the file need updating?
Whenever the pages it points to change in a meaningful way. A stale llms.txt file that links to moved or deleted pages is worse than none, because it sends a model to dead ends. We keep the file in sync as part of the ongoing engagement rather than generating it once and forgetting it.
