llms.txt is a proposed standard file, placed at the root of your site, that lists and describes your most important content to help AI systems find and use it. The honest answer to whether it works is that there is little evidence it currently has any meaningful effect: adoption is low and analyses have found AI systems largely ignore it. It is reasonable to add as zero-cost insurance, but it is not a priority and should not displace the work that matters.
In short
- llms.txt is a proposed file that curates your key content for AI systems.
- Current evidence shows little measurable effect: AI systems largely ignore it.
- One analysis found the overwhelming majority of valid llms.txt files received no requests.
- Add it as zero-cost insurance if you like, but prioritize crawler access, rendering, and structure instead.
What is llms.txt?
llms.txt is a proposed convention for a plain-text file at the root of your website, by analogy with robots.txt, that points AI systems to your most important content and provides clean, structured descriptions of it. The idea is to give AI models a curated map of your site so they can find and interpret your key pages more easily.
The proposal grew out of a reasonable concern. AI models work best with clean, well-structured text, and websites are often cluttered with navigation, scripts, and markup that make extraction harder, so the thinking was that a dedicated file listing your important pages in a tidy format could help AI systems use your content. As a concept it is sensible, and it is easy to create. The question that matters for marketers is not whether the idea is reasonable but whether AI systems actually use the file in practice, which is where the evidence comes in. The term itself is defined in the glossary entry for llms.txt.
Does llms.txt actually work?
The evidence indicates that llms.txt has little measurable effect today, because AI systems largely do not use it. Despite being widely discussed, the file does not appear to influence whether content is retrieved or cited in any meaningful way at present.
The data is fairly consistent on this. An analysis of a large set of domains found that the overwhelming majority of valid llms.txt files received no requests from AI crawlers at all over the period studied, meaning the files were simply not being fetched. Separate analyses of large domain samples found very low adoption and little sign that the file appeared among the sources AI systems actually cited. A senior figure at Google publicly compared llms.txt to the long-discontinued meta keywords tag, an element that search engines ignored because it was self-declared and easily gamed. The recurring theme is that a file in which a site describes its own important content, without external verification, is not something AI systems currently appear to rely on. This places llms.txt firmly in the low-impact category, alongside the contested evidence on schema covered in how structured data helps AI understand your site.
Should you implement llms.txt anyway?
You can implement llms.txt if you want to, since it costs almost nothing and does no harm, but you should not expect it to move your AI visibility, and you should not let it distract from higher-leverage work. Treat it as optional, low-priority insurance.
The reasoning is a simple cost-benefit. Creating an llms.txt file is quick and carries no downside, so adding one as a hedge, in case AI systems begin to use it more in future, is defensible. What is not defensible is treating it as an important AI-readiness step or spending significant effort perfecting it, because the current evidence does not support that. The risk is opportunity cost: time spent on llms.txt is time not spent on the things that demonstrably matter, namely crawler access, rendering, machine-readable structure, and the content and presence work. If you add it, add it quickly and move on.
What should you prioritize instead?
You should prioritize the foundations that the evidence actually supports: making sure AI crawlers can reach your content, that the content is visible without JavaScript, that it is well-structured, and that your brand reads as a clear entity. These are where AI-readiness is genuinely won.
The priority order is clear. First, crawler access, since AI systems cannot use content they cannot reach, covered in should you allow or block AI crawlers. Second, rendering, since many AI crawlers do not run JavaScript and client-side content can be invisible to them, covered in how to fix JavaScript rendering issues. Third, machine-readable structure, so content can be retrieved and used accurately, covered in how to make content machine-readable for LLMs. And fourth, entity clarity and the content and presence work covered in entity SEO for AI, AEO, and GEO. llms.txt sits well below all of these, which is why the honest recommendation is to handle the foundations first and treat llms.txt as an afterthought.