In its May 2026 llms.txt study, Ahrefs found that 97% of the roughly 38,000 domains with a valid file saw zero requests for it that month. (Ahrefs) So if you've been feeling guilty about not having one, you can relax. llms.txt is a plain-text Markdown file at the root of a website (yoursite.com/llms.txt) that gives AI systems a short, curated overview of the site and links to its most useful pages. It's a proposed convention, not a formal standard.
The short answer to "does it matter?" is: not for AI search visibility, based on everything published in the last year. The biggest search engine says its systems don't use it. Large crawl-log studies show AI search bots barely request it. Citation studies find no link between having one and being cited more.
There's one real exception, which is coding agents and developer docs. This page covers what the search engine said, what the data shows, who should still publish a file, and what to fix first if your actual goal is showing up in AI answers.
What Google says
First, the origin: llms.txt was proposed by Jeremy Howard in September 2024 as a convention, not a formal standard. (llmstxt.org)
Google's guide to its generative AI features has a "mythbusting" section, and llms.txt is in it. As SE Ranking summarizes it, Google Search does not use the file, and creating one will neither help nor hurt a site's visibility or rankings in Google Search, including its AI features. (Google Search Central; summary via SE Ranking)
The same guide says you don't need to "chunk" content or create special AI files to appear in AI Overviews or AI Mode. A page needs to be indexed and eligible to show in Search with a snippet, and there are no additional technical requirements. (Google Search Central)
That settles it for Google. Other assistants haven't published equivalent rules, so for ChatGPT, Claude and Perplexity the evidence comes from logs and citation data.
What the 2026 data shows
Three independent studies from the last year point the same way.
Most llms.txt files are never read
Ahrefs looked at all 137,210 domains in Ahrefs Web Analytics that received traffic in May 2026. 28% published an llms.txt file, and the 97% zero-request figure above comes from the subset whose file was valid. (Ahrefs)
Of the requests that did arrive, AI retrieval bots (the ones that fetch pages for live answers, such as OAI-SearchBot and PerplexityBot) accounted for just 1.1%. Ahrefs also found zero requests from AI bots for llms.txt files that don't exist, meaning no AI system goes looking for one. (Ahrefs)
No link to citations
SE Ranking analyzed close to 300,000 domains and found llms.txt on 10.13% of them. It built a model to predict how often a domain gets cited in LLM answers. Removing the llms.txt variable made the model more accurate, which means the file added noise rather than signal. (Search Engine Journal, November 2025; SE Ranking, September 2026)
The adoption numbers differ between the studies (28% vs about 10%) because the samples differ. Ahrefs' sample is sites using its analytics product, which skews toward SEO-aware teams.
Practitioners see the same thing in their logs
On X, the debate has mostly moved from "should we?" to "why are we still asking?". In August 2026, Aleyda Solis shared Mark Williams-Cook's write-up on the "cats.txt" test, describing the evidence offered for llms.txt as "GEO astrology" and walking through the faulty reasoning, starting with "It's definitely used, the LLM bots crawl it!". (Aleyda Solis on X) In April 2026, Jaan Porkon posted an analysis of 890 million web requests from March 2026 that found zero requests for /llms.txt from ChatGPT's, Claude's or Perplexity's bots. (Jaan Porkon on X) That's one dataset, but it matches the larger studies.
Why the myth persists
If the data is this one-sided, why do teams keep asking for it? A few reasons:
It's cheap. A file takes an hour, so it feels like a free bet.
Tools flag it. In May 2026, Chrome's Lighthouse added an llms.txt check to its Agentic Browsing audit category. (SE Ranking) A red mark in an audit looks like a problem even when it doesn't affect search.
Big AI companies publish one. OpenAI, Anthropic and Google's Gemini publish llms.txt files for their developer docs, and platforms like Wix, Mintlify, GitBook and Yoast generate them. (SE Ranking) Publishing a file for your docs is a different thing from AI search engines reading one.
The analogy to robots.txt and sitemaps sounds right. Those files work because crawlers agreed to read them. None of the major AI search products has said it reads llms.txt.
Who should still publish an llms.txt file
The one group with a real case is companies whose users work through coding agents. In Ahrefs' data, Claude Code, Anthropic's coding agent, out-fetched every AI retrieval bot, AI assistant and AI training crawler apart from GPTBot and one indexer. (Ahrefs) Ahrefs' conclusion: if your customers use coding agents, or agents act on your site, the file "stands a real chance of being read."
For a B2B SaaS company, that usually means:
Developer tools, APIs and SDKs with public docs.
AI and agentic products whose users build on them.
Any product where a developer is likely to point an agent at your docs and ask it to write an integration.
If that's you, put the file on your docs site and keep it current. Ahrefs notes that agents fetch llms.txt when directed, not speculatively, so link to it from your docs and mention it in your quickstart. (Ahrefs) If your product has no developer audience, you can skip it without losing anything measurable today.
If you publish one, keep it short and true
The format is simple: an H1 with the product name, a one-line summary in a blockquote, then sections of links with a short note on each. (llmstxt.org) For a B2B SaaS docs site, that could look like this:
# ExampleCo API
> ExampleCo is a billing API for B2B SaaS. This file lists the docs an agent needs to build an integration.
## Docs
- [Quickstart](https://docs.example.com/quickstart.md): create an API key and make a first request
- [Authentication](https://docs.example.com/auth.md): OAuth and API key options
- [Webhooks](https://docs.example.com/webhooks.md): event types and retry rules
## Optional
- [Changelog](https://docs.example.com/changelog.md)Two rules matter more than format. Every fact in the file has to match your docs and marketing pages, and the file needs an owner who updates it when the product changes. A stale llms.txt that agents do read is worse than none.
What to fix instead if you want AI visibility
If the real goal is getting named and cited in AI answers, the work that moves it is the same work described in the pillar on what AEO is. The short list, ordered by how often it's the blocker on B2B SaaS sites:
Let AI crawlers in. Check robots.txt and any bot-blocking in your CDN for OAI-SearchBot, PerplexityBot, ClaudeBot and Googlebot. Blocking them is a much bigger problem than a missing llms.txt.
Make key pages readable without JavaScript. Pricing, feature and comparison copy that only appears after scripts run may never reach a retrieval bot. The comparison of technical SEO agencies for SaaS covers AI crawler access and rendering in detail.
Check the Search Console setting. For Google's AI features, your site must be included in the "Search generative AI" setting. (Google Search Central)
Make your pages agree. A model can't be more accurate than the pages it reads, and conflicting pages produce confident, wrong summaries. That's the core of GEO.
Measure before and after. Track a fixed set of buyer prompts so you can see whether any change, llms.txt included, did anything. The guide to measuring AI search visibility sets this up.
For how this list differs from classic SEO work, see AEO vs SEO.
Verdict
llms.txt is a reasonable idea that the AI search engines haven't adopted. Google says it ignores the file, crawl logs show AI search bots almost never request it, and citation studies find no effect. The proposal is still evolving (SE Ranking notes a version 2 in August 2026), so this could change, and it's worth re-checking the data in a year.
Our recommendation is blunt. If developers point coding agents at your docs, publish a short, accurate file on the docs site and link to it. If not, don't bother, and be wary of anyone who sells llms.txt as an AI visibility fix. Spend that hour on crawler access, rendering and page consistency, which are the things AI answers actually depend on.
Crawler access and readability are covered in more depth on Miraa's technical SEO and LLM SEO pages.