Does llms.txt work?

llms.txt gets sold two ways. One says add it now or disappear from AI search. The other says it is pointless. The question buried under both is simple and answerable: does it actually work?

The honest answer is that it does not work as a ranking or citation lever, on current evidence. Google says it does not affect Search or AI Overviews, and studies find AI crawlers rarely fetch the file.

Where it does work is a narrower place: documentation and agents. This is what "work" would even mean for llms.txt, what the evidence shows, and whether it is still worth adding.

What "work" would even mean for llms.txt

Before judging whether it works, define the claim, because llms.txt is measured against a job it was never given.

The implied promise in most posts is that adding llms.txt improves your visibility in AI answers: more citations, better descriptions, an edge in AI Overviews. That is the claim to test. A separate, quieter claim is that it helps AI agents and tools navigate your site to complete a task. Those are different jobs, and llms.txt performs very differently at each.

Keep them apart and the confusion clears. The file mostly fails at the first job and genuinely helps with the second.

What the evidence says

On the visibility claim, the evidence is not encouraging, and it is worth being specific rather than vague.

Google has stated in its own guidance that llms.txt has no impact on Search rankings or AI Overviews, and its representatives have described the file as being about agent functionality rather than search discovery. That covers the single largest AI surface your buyers touch.

Independent research across large numbers of domains has found that AI crawlers rarely fetch the file at all. So the mechanism by which llms.txt is supposed to help, a model reading it and acting on it, largely is not happening yet. A file that is not read cannot influence what a model says about you. On the visibility question, that is close to a settled answer for now.

Where llms.txt does work

The second job is where the file earns its place, and it is real.

AI coding assistants and agents working with a product's documentation benefit from a short, curated map of the technical pages. It helps them find the correct reference quickly and complete a task accurately, which is a genuine, active use case rather than a hypothetical one. If you run developer docs, a knowledge base, or a product that agents interact with, llms.txt does something for you today.

For a typical marketing or consulting site, that use case rarely applies, which is why the honest verdict for most readers is "harmless, not helpful."

Why add it anyway

Given all that, there is still a reasonable case for adding it, as long as you are clear about what you are buying.

It is cheap and low risk, and it positions you if adoption grows and the major systems start reading it. That is a small bet on a possible future, not a present-day fix. The right way to hold it is: add the file, generate it so it stays current, then forget about it and spend your attention elsewhere. A worked llms.txt example shows how little effort a good one takes, and what is llms.txt covers the format and the "do you need it" question in full.

The mistake is not adding llms.txt. The mistake is believing it works as a visibility lever and neglecting the things that do.

The comparison worth making

llms.txt is not the first file sold as an AI visibility fix, and the pattern is instructive.

Schema markup gets the same treatment: added in bulk on the promise that it will lift AI citation, when its real job is comprehension, not ranking. The parallel is exact enough that why schema markup is not helping your AI visibility reads as the same lesson in a different file. Both help a machine understand your site. Neither is the reason a machine cites you.

What earns citation is upstream of both: content a crawler can reach and render, answers written to extract cleanly, a consistent identity, and material a model cannot synthesise without you. Files at your domain root are housekeeping around that work, not a substitute for it.

The takeaway

Does llms.txt work? Not as an AI ranking or citation lever, on today's evidence: Google does not use it for Search or AI Overviews, and crawlers rarely read it. It does work in documentation and agent settings. Add it if you like, as a cheap bet on future adoption, but do not mistake it for the work that actually earns AI visibility.

If you want to know what is genuinely driving or blocking your visibility in AI answers, rather than guessing at files, that is what an AI visibility audit provides.

This article is part of the SEO in the AI Era: The Complete Guide guide.