What Is llms.txt, and Should You Add One?
An emerging convention with no guaranteed payoff and a very low cost. Here's an honest read on whether it's worth your time.
llms.txt is a plain-text file at the root of your domain — yoursite.com/llms.txt — that describes what your site is and points to the pages that matter. Think of it as a README written for a language model rather than a developer.
Why it was proposed
Language models work within a context window. When one encounters your site, it may pull a page or two, and those pages carry navigation, cookie notices, footers and markup that have nothing to do with your actual subject. A model can waste most of its attention on furniture.
llms.txt proposes a shortcut: one curated file, in Markdown, saying what you do and which URLs are worth reading. No parsing, no guessing.
Let's be honest about status
This is a convention, not a standard. No major AI company has committed to reading it. There is no published evidence that adding one changes how often you get cited. Anyone telling you otherwise is guessing.
So why does our audit check for it? Because the cost is ten minutes and the downside is zero, while the upside — if adoption comes — is being early on a file that describes your site in your own words. That is a reasonable bet, not a guarantee, and we score it as a small signal rather than a major one.
What it looks like
# AnswerRank
> A free tool that audits whether AI answer engines can read,
> understand and cite a website, and generates the exact fixes.
## What the tool does
- Scores 19 AI-visibility signals and 12 technical health checks
- Detects JavaScript-rendered content invisible to AI crawlers
- Checks robots.txt access for GPTBot, ClaudeBot, PerplexityBot and others
## Key pages
- [Free audit](https://answerrank.arisesaga.com/): run a report on any URL
- [Guides](https://answerrank.arisesaga.com/guides): how AI search visibility works
- [About](https://answerrank.arisesaga.com/about): how the engine works
## Contact
Sastihari S — suryapriya.s@finstein.aiWriting a good one
Lead with what you are. The first two lines should let a reader who has never heard of you say what you do. Skip the marketing adjectives; models don't weight enthusiasm.
Link the pages you'd want quoted. Not your whole sitemap — that already exists. Ten to twenty URLs that genuinely represent you.
Use plain Markdown. Headings, bullets, links. No HTML, no tables, no cleverness.
Keep it current. A file describing last year's product is worse than none, because it is confidently wrong.
Where it goes
The web root, served as text/plain, reachable at /llms.txt. On Next.js put it in public/. One trap worth knowing: single-page apps with catch-all routing often return the HTML shell for any unknown path, so /llms.txt appears to exist but serves a web page. Our audit checks for that specifically — if the response starts with <, it doesn't count.
What it will not do
It will not make a blocked crawler read you — robots.txt decides that. It will not compensate for content that isn't in your HTML. It will not manufacture authority. It is a small, cheap, optional signal, and worth exactly that much.
If you have the ten minutes, do it. If you are choosing between this and fixing a client-rendered page or an accidentally blocked crawler, those come first, every time.
Run a free audit to see whether your site publishes one.