Does llms.txt do anything?
Over 844,000 sites have adopted llms.txt. No major AI provider has committed to reading it in production retrieval, and the largest correlation study found no relationship with citations. Here is what it is actually for.
llms.txt is not adopted by any major AI provider as a discovery or ranking mechanism. Selling it as an AI visibility lever is not defensible. It has a genuine use, and that use is something else entirely.
llms.txt is a plain text file at the root of a website, written in Markdown, describing what the site is and where its important pages live. The proposition is that AI assistants will read it and understand your business better, and therefore recommend you more often. It is one of the most widely adopted ideas in AI visibility, and one of the least supported.
What the providers actually say
- Google: John Mueller has confirmed that no Google Search system reads or acts on llms.txt. Google's own May 2026 guidance states you do not need to create new machine readable files, AI text files, markup, or Markdown.
- OpenAI: the documented crawler-control recommendation is robots.txt. Server-log analyses consistently show GPTBot fetching llms.txt occasionally, and rarely.
- As of Q1 2026, no major AI company — OpenAI, Google, Anthropic, Meta or Mistral — has publicly committed to reading or acting on llms.txt in production retrieval.
What the measurement says
SE Ranking examined 300,000 domains and found no correlation between AI citations and the presence of an llms.txt file. Search Engine Land found no effect across a smaller sample. More than 844,000 sites have adopted it on zero proven impact.
Note what that does and does not mean. No correlation across 300,000 domains is reasonably strong evidence of no large effect. It is not proof of no effect at all, and llms.txt is too young for anyone to be certain. But the burden of proof sits with the people charging for it.
What about Shopify rolling it out?
Shopify has begun serving an llms.txt file on stores, and some also expose a sitemap_agentic_discovery.xml. Given Shopify powers millions of stores, that is a very large jump in adoption, and it is being reported in this category as evidence that the format matters.
Be careful about what it demonstrates. A platform generating a file for its merchants is evidence of platform adoption. It is not evidence that any assistant reads the file or weights it, and Shopify has not published full documentation on the rollout. Adoption and effect are separate questions, and this category routinely reports the first as though it settled the second.
It is still worth looking at yours if you run a store, for a different reason: it shows you how your products, collections and policies are being summarised. That is a useful audit of your own data. It is not a visibility lever.
The legitimate use, which is not search
llms.txt has real traction as a convention for developer documentation. Anthropic recommends it in its guidance on writing for agents, and Stripe, Cursor, Cloudflare, Vercel, Mintlify and Supabase all publish one.
That is a developer-experience use case: making technical documentation consumable by coding agents that have been pointed at your docs deliberately. It is not a mechanism by which a consumer assistant discovers and recommends a local business.
Conflating those two things is the single most common piece of misinformation in this category. Citing Anthropic's support for llms.txt as evidence that it affects ChatGPT recommendations is a misrepresentation, whether or not the person doing it knows.
So should you have one?
It takes about twenty minutes and it harms nothing. If you want one, have one. The test is what you are told it does and what you are charged for it.
- Being told it is good practice and costs an hour: reasonable.
- Being told it will get you recommended by ChatGPT: not supported by any provider statement or study.
- Being charged a monthly retainer for maintaining it: no.
What we did on our own site
We publish one at sentinay.com/llms.txt. We added it because it is standard, correct, and costs nothing to keep accurate, and because a business selling this work looks unserious without the basics in place. We do not expect it to change how often we get named, we do not charge for it, and if it does nothing we will say so here.
That is the general shape of the honest answer for most of the cheap technical items in AI visibility. Do them because they are correct. Do not buy them as growth.
Common questions
- Do I need an llms.txt file?
- No. No major AI provider has committed to reading it in production retrieval, and the largest study found no correlation between llms.txt and citations across 300,000 domains. It takes about twenty minutes and harms nothing, so have one if you want one, but it should not be a line item.
- Does Google use llms.txt?
- No. Google has confirmed that no Google Search system reads or acts on it, and its guidance states you do not need to create AI text files or Markdown for AI features.
- Why do so many agencies recommend llms.txt then?
- It is cheap to deliver, easy to show a client, and sounds specific to AI. Anthropic recommends it for developer documentation consumed by coding agents, and that recommendation is often presented as evidence that it affects consumer assistant recommendations. It is not the same thing.
Sources
Every claim above should be checkable. Where a study has limits, they are stated rather than left out.
- Google Search Central — May 2026 guidance on AI features and machine-readable filesGrade A. Provider documentation.
- SE Ranking — llms.txt and AI citations across 300,000 domainsGrade B. No correlation found.
- Anthropic — writing documentation for agentsGrade A for the developer-documentation use case. Not a search claim.
- OpenAI — crawler and robots.txt documentationGrade A. Provider documentation.
More on getting cited in ai answers
- Why the five AI assistants disagree about who to recommend
We measured every source five assistants used to answer the same questions. Not one domain was cited by all five, and 89% were cited by only one. Our own original data on why a single AI visibility number is misleading.
- The source ecosystem: how AI actually decides who to name
When we asked five assistants who to hire in our own category, 59 of 108 citations were other people's roundup lists. Five of the twelve most-named companies sat on one directory. Being named is mostly about other people's pages.
- What actually works in AI visibility, graded by evidence
Twenty interventions sold as AI visibility work, each rated by how strong the evidence behind it actually is. Three are gates you must pass. One has a rigorous causal study behind it, and that study found nothing.
- How AI assistants pick which businesses to name
ChatGPT, Perplexity, Google's AI answers and Copilot use four different pipelines to decide who gets recommended. An intervention that works on one can do nothing on another, which is why they have to be measured separately.
- What to fix first for AI search
A running order based on what the evidence supports rather than what is easiest to sell. Three gates before anything else, then off-site work, then your own pages — which come later than almost every agency will tell you.
Want to know where you stand?
We ask five AI assistants for a company like yours and send you a free report showing how often you came up and who came up instead.
Get my free report