RankEcho
Home / Research / How many websites have an llms.txt? What 2,000 sites show
AI Search Intelligence

How many websites have an llms.txt? What 2,000 sites show

The short answer

More than you might think in marketing and software, few elsewhere. In November 2026, 57.4% of 1,000 sites cited in AI answers in RankEcho audits, mostly marketing, SEO and software sites, served an llms.txt, and 54.7% opened it with the Markdown heading the proposal asks for. Among 1,000 small-business sites it was 22.2%. The files are small: a median of 45 links on the cited sites and 11 on small-business sites. AI crawlers still rarely ask for them: on rankecho.io they fetched llms.txt 17 times in 30 days and robots.txt 394 times.

Key findings

Two cohorts of 1,000 sites each, read on 16 and 17 November 2026, the same as our report on AI crawler blocking: the hosts AI engines cited most in RankEcho audit answers, which are mostly marketing, SEO and software sites, and small-business homepages from public business listings. Shares are of the sites that answered: 988 and 990.

  • Adoption splits by market. 567 of the cited sites (57.4%) serve an llms.txt; 220 small-business sites (22.2%) do.
  • Most files follow the proposal's shape. 497 of the 567 files on cited sites (88%) open with a Markdown H1 and list at least one link; 181 of 220 (82%) on small-business sites.
  • They are short lists. The median file links 45 pages on the cited sites and 11 on small-business sites, about 11.5 KB and 3.5 KB of text.
  • Some files list nothing. 48 files on cited sites and 22 on small-business sites have no Markdown links at all.
  • Some sites send mixed signals. 48 cited sites publish an llms.txt for language models while their robots.txt blocks a training crawler such as GPTBot. Only 2 also block an AI search or user-fetch agent.
  • AI crawlers rarely read the file. On rankecho.io, 17 October to 15 November 2026, the AI crawlers that fetched llms.txt did so 17 times in all (ClaudeBot 10, Applebot 3, OAI-SearchBot 1, ChatGPT-User 1, Amazonbot 1, Meta-ExternalAgent 1), and fetched robots.txt 394 times.

Who publishes an llms.txt

llms.txt is a proposed Markdown file at a site's root: a heading with the site's name, a short summary, then lists of links to the pages a language model should read. Nothing requires it, and Google has said it does not use the file for Search. Adoption is a choice each site makes.

The gap between the two cohorts is the story. Marketing, SEO and AI-visibility software sites, which make up most of the cited cohort, have taken the file up quickly; many of the tools in that market recommend it. Small businesses mostly have not. This data cannot say who wrote each file; some SEO plugins and site builders can now generate one.

Share of sites with an llms.txt, and how many follow the proposal's shape (cited n=988, small business n=990)
Serves an llms.txt57.4%22.2%
Opens with a Markdown H154.7%20.0%
H1 and at least one link50.3%18.3%
Cited sitesSmall-business sites
Serves an llms.txt567 (57.4%)220 (22.2%)
Opens with a Markdown H1540 (54.7%)198 (20.0%)
H1 and at least one link497 (50.3%)181 (18.3%)

What the files look like

On the cited sites, most files list 10 to 99 links; on small-business sites a third list fewer than ten. 194 files on cited sites and 36 on small-business sites were longer than the 20,000 characters we read, so their link counts here are lower bounds.

Share of llms.txt files by number of Markdown links
No links8.5%10.0%
1 to 9 links6.0%35.9%
10 to 99 links75.5%49.1%
100 or more links10.1%5.0%
Links in the fileCited sitesSmall-business sites
None4822
1 to 93479
10 to 99428108
100 or more5711
Median links4511
Median size11.5 KB3.5 KB

Publishing the file versus crawlers reading it

Adoption measures what sites publish, not what AI engines use. rankecho.io serves an llms.txt too, and its Cloudflare logs show how often AI crawlers ask for it. From 17 October to 15 November 2026, the AI crawlers that fetched it did so 17 times in all: ClaudeBot 10, Applebot 3, OAI-SearchBot 1, ChatGPT-User 1, Amazonbot 1, Meta-ExternalAgent 1. The same crawlers fetched robots.txt 394 times.

ClaudeBot fetched the file more often than in our first 30-day count, so this is worth watching. But the pages an llms.txt lists are found through links and sitemaps either way, and no AI engine has said it ranks or cites pages because of the file.

Should you publish one?

It costs little, and the market you sell into may expect it. If you publish one, make it accurate: a heading with your name, one plain summary line, and links to the canonical pages you would want an assistant to read first. Keep private and redirected URLs out, and update it when pages move.

Do the work crawlers demonstrably use first: a robots.txt that lets AI search crawlers in, a complete sitemap, and pages whose content is in the HTML.

Method

Each host's /llms.txt was fetched once, with up to three attempts, between 16 November 12:00 and 17 November 00:01 UTC, with the user agent RankEcho-AgentCheck/1.0, alongside its robots.txt. A file counts when it answered with a 2xx status and a non-empty body that is not an HTML page. Its first non-blank line counts as an H1 when it is a Markdown "# " heading; links are Markdown links. At most 20,000 characters were read.

The cited cohort is the 1,000 hosts AI engines cited most in RankEcho audit answers. Most of those answers came from audits of rankecho.io, whose prompts ask about AI visibility and SEO software, so the cohort is mostly marketing, SEO and software sites. The small-business cohort is 1,000 homepages from the public US business listings behind our website readiness screening. No domain is named here; only aggregate counts are published.

Limits

A site can serve a plain-text answer at /llms.txt that is not really an llms.txt, such as a catch-all text page; the H1 and link counts are the stricter reading. Files longer than the read limit were cut, so very large files are undercounted.

The cited cohort describes one market, not the web at large, and the small-business cohort comes from US listings. Both were read once, and llms.txt files change. The crawler counts come from one site's logs.

Published counts of the most-visited or most-linked sites put adoption at roughly 5% to 12%. Samples and definitions differ between studies, so compare shares with care.

Frequently asked questions

How many websites have an llms.txt?

In our November 2026 sample, 57.4% of 1,000 sites cited in AI answers (mostly marketing, SEO and software sites) and 22.2% of 1,000 small-business sites served one.

Do llms.txt files follow the proposal?

Mostly. 88% of the files on cited sites and 82% on small-business sites open with a Markdown heading and list at least one link.

How long is a typical llms.txt?

The median file lists 45 links on the cited sites and 11 on small-business sites, about 11.5 KB and 3.5 KB.

Do AI crawlers read llms.txt?

Rarely, on our own site. From 17 October to 15 November 2026, AI crawlers fetched rankecho.io's llms.txt 17 times and its robots.txt 394 times.

Can I use this data?

Yes. The aggregate table is a CSV under CC BY 4.0; please link to this page. No domains are published.

See where AI ignores your brand — run a free audit →
Last updated 2026-11-17 · RankEcho · Operated by Nexus Decision Systems LLC