A practical look at llms.txt implementation: what the file proposes, current adoption reality, how to write one, and where effort is better spent first.
llms.txt implementation has become a common item on AI visibility checklists. Before adding it, it is worth being clear about what the file is, what it is not, and how much of its promise is currently realized.
What the proposal is
llms.txt is a proposed convention: a markdown file at your domain root that gives language models a curated, clean map of your most important content. The reasoning is that models parsing a large marketing site waste context on navigation, scripts, and boilerplate, and may miss the pages that matter.
The file lists key pages with short descriptions, optionally pointing to plain-text or markdown versions of each. It is a hint offered to consumers of your content, similar in spirit to a sitemap but aimed at comprehension rather than discovery.
Note what it is not. It is not an access control mechanism, it is not a ranking factor, and it does not compel any system to read it.
Adoption reality
Support is uneven. The convention has real traction among documentation-heavy and developer-facing sites, and tooling has grown around it. Adoption by the major assistants as a primary retrieval path is far less established than checklist articles imply.
The honest position: the cost of publishing one is low, the downside is negligible, and the demonstrated benefit is modest and concentrated in documentation use cases. Treat it as cheap insurance rather than a lever.
How to write one
Keep it short and curated. The value comes from selection, not completeness.
Open with an H1 of your organization name, followed by a blockquote of one or two sentences describing what you do. Then group links under H2 headings by purpose, each link followed by a colon and a brief description of what the page covers.
Prioritize documentation, service definitions, pricing or engagement models, and any original data you publish. Exclude navigation pages, tag archives, and thin content.
If you maintain plain markdown versions of key pages, link those rather than the rendered HTML, since they parse more cleanly.
Keep it under a few dozen links. A file listing three hundred pages recreates the problem it was meant to solve.
Where to spend effort first
If your AI visibility budget is limited, llms.txt is not the first item.
Verify AI crawler access at robots and CDN level, since blocking makes everything else moot. Confirm your content renders server-side. Correct inaccurate descriptions of your business wherever they appear. Restructure your highest-value pages so key claims are self-contained and quotable. Build corroboration off-site.
Each of those has clearer demonstrated impact than the file. Publish llms.txt after them, not instead of them.
Maintenance
A stale llms.txt is worse than none, because it actively points at outdated material. Regenerate it from the same source that drives your sitemap, or set a quarterly review. If nobody owns it, do not publish it.
Frequently asked questions
Does llms.txt control whether models train on our content?
No. Training and crawling permissions are governed by robots directives, terms, and licensing arrangements, not by this file.
Will it improve our citation rate?
There is limited evidence of a direct effect for general marketing sites. Documentation-heavy sites report better outcomes, likely because their content suits the format.
Where does the file go?
At the root of the domain, served as plain text or markdown, at the conventional path.
Should we also publish markdown versions of pages?
For documentation, often worthwhile. For general marketing content, the maintenance cost usually exceeds the benefit.
Explore our software and web capabilities or contact Voquarn Code to prioritize AI visibility work by actual impact.
Written by
Moueen Togarvi
Founder & CEO at Voquarn Code, focused on product engineering, search growth, and practical AI systems.