Do you need an llms.txt file?
No major engine has committed to reading llms.txt, so treat it as cheap insurance rather than a lever. The better argument for publishing one is that writing it forces you to decide which twenty pages actually represent you, and that decision usually improves everything else you do.
What is llms.txt?
A plain-text file at your site root, written in markdown, offering models a curated map of your most useful pages. The idea borrows from robots.txt and sitemap.xml.
Will it improve my AI visibility?
There is no evidence that it will on its own, and no major engine guarantees support. Anyone promising a visibility gain from publishing one is ahead of the evidence.
The cost is an afternoon, which is why publishing one is still reasonable. The mistake is treating it as a strategy rather than housekeeping.
If you publish one, make it useful
A list of every URL is a sitemap. The value here is the twenty pages you would want quoted.
One line saying what the page authoritatively covers, so a model can choose between them.
Hand-maintained files drift within a quarter and then quietly misdescribe you.
Ours says we publish no customer results, so a model cannot invent one on our behalf.
Common questions
No. Robots controls access, llms.txt suggests priority. They answer different questions and you want both.
Writes up what the stored answers show, one category at a time.