llms.txt Index · run 1 · read 24-25 Oct 2026
EU llms.txt Index: how many popular EU sites publish an llms.txt
On 24-25 October 2026, 11.7% of 10,223 popular EU websites published an llms.txt file; 79.2% returned 'not found'. The sites are the top domains under 25 country-code TLDs, Tranco list 56Z8N (CitedBy EU llms.txt Index). Romanian-domain sites are most likely to have one (21.0%), French-domain sites least (4.9%). Sites that ask GPTBot to stay out of robots.txt are less than half as likely to publish an llms.txt (5.7% against 12.8%). WordPress sites publish an llms.txt twice as often as Drupal sites (15.2% against 7.7%); many WordPress files are written by SEO plugins (Yoast, Rank Math, All in One SEO) rather than by hand.
Run 1 · /llms.txt requested 24-25 Oct 2026 · Tranco list 56Z8N · 10,223 of 10,426 sites answered · Archive of run 1. Download the per-site CSV
What this measures
This index records whether a site publishes a file at /llms.txt, the Markdown file that llmstxt.org proposes as a guide for AI tools. It measures presence, not use: it does not show whether any AI engine reads the file, and it does not measure what any assistant cites. The same sites, from the same sample, are read for robots.txt in the EU AI-Access Index; to draft an llms.txt for your own domain, the free llms.txt generator builds one from your home page and sitemap.
| Result of the /llms.txt request | Sites | Share of the 10,223 that answered (95% interval) |
|---|---|---|
| Have an llms.txt (HTTP 200, a text file with a Markdown title or link) | 1,200 | 11.7% (11.1 to 12.4%) |
| Answer "not found" | 8,098 | 79.2% (78.4 to 80.0%) |
| Answer 200 with a web page or text that is not an llms.txt (not counted) | 925 | 9.0% (8.5 to 9.6%) |
203 sites did not answer at all and are outside the share base. Intervals are Wilson 95% intervals. 925 sites answered 200 with a page that is not an llms.txt (a soft-404 or an error text): they are not counted as having one.
By country-code domain
| Country (ccTLD) | Sites answered | With an llms.txt | Share (95% interval) |
|---|---|---|---|
| Romania (.ro) | 420 | 88 | 21.0% (17.3 to 25.1%) |
| Estonia (.ee) | 438 | 77 | 17.6% (14.3 to 21.4%) |
| Lithuania (.lt) | 407 | 65 | 16.0% (12.7 to 19.8%) |
| Denmark (.dk) | 427 | 68 | 15.9% (12.8 to 19.7%) |
| Greece (.gr) | 406 | 57 | 14.0% (11.0 to 17.8%) |
| Slovenia (.si) | 404 | 56 | 13.9% (10.8 to 17.6%) |
| Ireland (.ie) | 419 | 58 | 13.8% (10.9 to 17.5%) |
| Luxembourg (.lu) | 156 | 20 | 12.8% (8.5 to 19.0%) |
| Portugal (.pt) | 410 | 52 | 12.7% (9.8 to 16.3%) |
| Netherlands (.nl) | 425 | 53 | 12.5% (9.7 to 16.0%) |
| Croatia (.hr) | 422 | 51 | 12.1% (9.3 to 15.5%) |
| Latvia (.lv) | 383 | 44 | 11.5% (8.7 to 15.1%) |
| Slovakia (.sk) | 413 | 46 | 11.1% (8.5 to 14.5%) |
| Belgium (.be) | 407 | 45 | 11.1% (8.4 to 14.5%) |
| Finland (.fi) | 390 | 41 | 10.5% (7.8 to 14.0%) |
| Bulgaria (.bg) | 422 | 44 | 10.4% (7.9 to 13.7%) |
| Poland (.pl) | 420 | 43 | 10.2% (7.7 to 13.5%) |
| Sweden (.se) | 447 | 45 | 10.1% (7.6 to 13.2%) |
| Czechia (.cz) | 454 | 45 | 9.9% (7.5 to 13.0%) |
| Spain (.es) | 418 | 41 | 9.8% (7.3 to 13.0%) |
| Italy (.it) | 421 | 39 | 9.3% (6.9 to 12.4%) |
| Germany (.de) | 452 | 40 | 8.8% (6.6 to 11.8%) |
| Austria (.at) | 402 | 30 | 7.5% (5.3 to 10.5%) |
| Hungary (.hu) | 434 | 31 | 7.1% (5.1 to 10.0%) |
| France (.fr) | 426 | 21 | 4.9% (3.2 to 7.4%) |
Sorted by share. Luxembourg has 156 sites answered; its interval is wide. Download by country (CSV)
By CMS
| CMS (home-page signature) | Sites answered | With an llms.txt | Share (95% interval) |
|---|---|---|---|
| No signature matched | 5,518 | 483 | 8.8% (8.0 to 9.5%) |
| WordPress | 1,632 | 248 | 15.2% (13.5 to 17.0%) |
| Home page not fetched | 1,401 | 264 | 18.8% (16.9 to 21.0%) |
| Next.js | 579 | 91 | 15.7% (13.0 to 18.9%) |
| Drupal | 518 | 40 | 7.7% (5.7 to 10.3%) |
| Magento | 205 | 43 | 21.0% (16.0 to 27.1%) |
| PrestaShop | 140 | 10 | 7.1% (3.9 to 12.6%) |
| TYPO3 | 129 | 10 | 7.8% (4.3 to 13.7%) |
| Joomla | 60 | 1 | 1.7% (0.3 to 8.9%) |
Only CMSs with at least 50 sites are shown; the others have fewer and are not reported. The CMS is a signature in the home page, not a verified platform, and "No signature matched" is the largest group. Many WordPress files are written by SEO plugins (Yoast, Rank Math, All in One SEO): a plugin header reading "Generated by … this is an llms.txt file" was seen among the 74 files that do not open with a title line; we did not count it in the rest. Download by CMS (CSV)
Next to the robots.txt result
| GPTBot in robots.txt (AI-Access Index run 1) | Sites answered | With an llms.txt | Share |
|---|---|---|---|
| Not blocked | 8,649 | 1,110 | 12.8% |
| Blocked | 1,574 | 90 | 5.7% |
A pattern across sites, not a cause: this index does not show why a site did or did not publish a file.
Method and limits
- Question: how many popular EU websites publish a real
/llms.txt, the Markdown file at the site root that llmstxt.org proposes as a guide for AI tools? - Sample: the 10,426 sites from the AI-Access Index run 1 whose robots.txt was read (top 600 domains by Tranco rank under each of 25 EU ccTLDs, Luxembourg 250; Tranco list 56Z8N, created 22 Oct 2026; tranco-list.eu/list/56Z8N/1000000). Reading robots.txt first means every site counted was alive. 10,223 answered the
/llms.txtrequest (203 unreachable). - Fetch: one GET of
/llms.txtper site, first 100 KB, https then www then http; a further address is tried only when the earlier one gave no answer at all (so one request per site, as the User-Agent says:CitedBy-research/1.0 (+https://citedby.eu; llms.txt study, one request per site)). Redirects were followed. No login, no cookies, no browser identity. - Reading: "has an llms.txt" = HTTP 200, not HTML (a leading BOM is ignored), and the text has a Markdown H1 line or a Markdown link. A 200 answer that is a web page, or text with neither, is a soft-404 and is not counted (925 sites, 9.0%).
- Rule change, said openly: the first pass counted only files whose first line is an H1 (948 files). A 40-site check of the rejected 200-answers found 10 plain-text files, 8 of them real llms.txt (a BOM, or a plugin header "Generated by Yoast SEO / Rank Math / All in One SEO ... this is an llms.txt file"). The rule was widened and only the 1,213 rejected sites were probed again. Result 1,200 files; 1,126 (93.8%) satisfy the strict H1 form. The strict flag is in the CSV (
strict_h1). - Precision check (25 Oct): 15 "valid" sites drawn at random and re-fetched: 15 real llms.txt (a title, a one-line description and links). 12 of the files that do not open with an H1 re-fetched: 11 plugin files, 1 had changed to an HTML page between the two reads, so a site's answer can differ between reads.
- Parser test: 8 unit tests on the reading rule.
- Limits: Presence is measured, not use: a file existing says nothing about whether any AI engine reads it. ccTLD is not the operator's country. Tranco ranks popularity, so this is the popular end of each ccTLD, not a random draw of all sites. 2% of sites did not answer. CMS is a home-page signature (largest group "no signature matched"). A site can serve different answers on different days (one of 12 re-reads changed). Plugin defaults count: a generated file is a file. We did not judge the quality of any file.
- Raw data: one row per domain with the llms.txt result, public data only (whether a public file exists): llms-txt-index-2026-10.csv (10,426 rows). Licence: CC BY 4.0, cite as "CitedBy EU llms.txt Index, run 1, October 2026".
Run history
The crawl is repeated on the newest Tranco list. This page is run 1 (read 24-25 Oct 2026) and stays at /data/llms-txt-index-2026-10; /data/llms-txt-index always shows the latest run.
Does your site answer /llms.txt?
The free checker shows whether your /llms.txt is present, valid, working and consistent with your robots.txt and sitemap. With the paid audit, we crawl your site up to your plan's page limit and give you the exact fix for each issue. From €9, 14-day money-back guarantee.
Checked on 25 Oct 2026 · Source: Data (CSV) · Method: How CitedBy checks · Change log: What changed