Is an llms.txt file worth adding to my site?
It is cheap and harmless, and there is no public evidence that major assistants read it. llms.txt is a proposed convention - a markdown file listing your key pages and what they cover. No major provider has committed to consuming it. Add it if you want the tidiness, but do not treat it as a substitute for crawlable pages, clean structure and consistent facts, which are the things that demonstrably matter.
What it is, and what it is not
The idea is a plain markdown file at a known path that summarizes your site and points to the pages most worth reading, so a language model can orient itself without crawling everything. In spirit it sits somewhere between robots.txt and a sitemap.
It is a proposal, not a standard. There is no governing body, no conformance test, and no published commitment from the large assistant providers to fetch or honor it.
Why it has not been adopted the way sitemaps were
Sitemaps succeeded because the party that wanted the file was the same party that consumed it, and the file solved a real discovery problem for a crawler that already existed.
llms.txt asks a consumer to trust a self-authored summary of a site over the site itself. A retrieval system that already crawls the page has little reason to prefer your description of it. That is not a criticism of the idea so much as an explanation of the slow uptake.
It may still get adopted. Conventions sometimes win because enough sites publish them that consuming them becomes cheap. That is a reason to treat the file as a low-cost option rather than as a mistake - just not as a strategy.
If you publish one, publish it honestly
Keep it short. List real URLs with one accurate line each. Keep it in sync when pages change, because a stale index of your site is another conflicting source about your business, which is the specific problem you are trying to avoid.
Resist the urge to stuff it with claims and keywords. If any system ever does consume these files, the ones that read like marketing copy will be the first discounted.
Treat it the way you would treat a sitemap: generated from the same source as your navigation, not maintained by hand in a text editor by whoever remembers it exists.
What to do with the same hour instead
The work that has observable effects is less exciting. Confirm your important pages are in the initial HTML. Rewrite the first paragraph of your top pages so each answers its question outright. Reconcile hours, address and services across your site, your profile and the aggregators.
Any one of those does more than a speculative file. If you want a machine-readable summary of your expertise that systems already know how to read, structured data and a well-organized answer hub are the versions that exist today, which is where we put the effort in AI search work.
The broader pattern is worth naming, because this category produces a new must-do file every few months. Before adopting one, ask who consumes it, whether they have said so publicly, and what observable thing would change if you were right.
Topics: llms.txt · standards · myths · technical
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.