Most .jp sites that "have" llms.txt send a web page

We asked Japan's busiest sites for one file

On 26 October 2026 our weekly scanner visited the most visited .jp sites and asked each for the same file: /llms.txt. About 1 in 8 answered "200 OK", which means "here it is". But when we read what came back, most of those yeses were not the file. They were a web page.

What is llms.txt, and why are you asked about it?

llms.txt is a plain-text file at the root of a site that summarises the site for AI apps. It is a proposal, not a standard, and this post does not claim that any AI app reads it. Still, many web leads in Japan have been asked: "Do we have llms.txt? Do our competitors?"

The quickest way to answer sounds reasonable. Type in the file's address, see whether the server says "200 OK" (yes) or "404" (no), and count.

What we saw when we read the answers

Among the sites that said "200 OK", about 7 in 10 sent back a web page instead of a text file. Web people call this a soft 404: the server says "yes", but the thing you asked for is not there.

• A real file29 sites29.6%
• A web page (soft 404)69 sites70.4%

Source: our weekly scan of large .jp sites (Tranco list of 25 October 2026), one request per site from a Japanese IP under our scanner's name. 26 October 2026: 98 of 777 sites answered "200 OK". All 69 web pages came back as text/html.

We only counted a real file when the answer was plain text or Markdown, not empty, and did not start with an HTML tag. Counted by "200 OK" alone, llms.txt looks more than three times as common as it is.

Ask the site for /llms.txt

200

"200 OK"The server says yes; not proof on its own

Read the answerPlain text or Markdown, not empty, no HTML tag at the start?

• yes
  If yes: a real file
• no (the other branch)
  If no: a soft 404A web page answering for a file that is not there

Source: our weekly scan of large .jp sites (Tranco list of 25 October 2026), one request per site from a Japanese IP under our scanner's name. The filter does not check the file's syntax.

Not a one-week blip

We scanned the same list again on the next two Mondays. Each time it was the same: about twice as many sites sent a web page as had a real file, and the number of real files barely moved.

Web page with "200 OK"Real file

	Web page with "200 OK"	Real file
26 Oct	69	29
2 Nov	66	30
9 Nov	63	30

Source: our weekly scan of large .jp sites (Tranco list of 25 October 2026), one request per site from a Japanese IP under our scanner's name. Sites checked: 777 on 26 October, 782 on 2 November, 775 on 9 November 2026.

What does this mean for you?

When you read a report on what share of Japanese sites have llms.txt, ask how it counted. We do not know how other surveys count. We only see that the way of counting can triple the number.

Counted by "200 OK"		98 of 777 (12.6%)
Counted as a real file		29 of 777 (3.7%)

Source: our weekly scan of large .jp sites (Tranco list of 25 October 2026), one request per site from a Japanese IP under our scanner's name. 26 October 2026, 777 sites whose answer stayed on their own domain. Other sites may differ.

Your own server may be doing the same. If your server or site builder returns a web page with "200 OK" for any address that does not exist, every checker that only looks at the status will tell you that you have llms.txt, even if you never made one. The same goes for your competitors: "they already have llms.txt" may just be a web page answering in its place.

How do you check in a minute?

1. Open `yourdomain/llms.txt` in a browser. If you see a styled web page instead of plain text, it is most likely a soft 404.
2. If you use the command line, run curl -I https://yourdomain/llms.txt and look at the status and the Content-Type line. With no file, the right answer is 404. With a file, the type should be plain text or Markdown.
3. If you get "200 OK" and a web page, ask whoever runs your server to make unknown addresses return 404.
4. Or use the free check on `pw` (our research site). It reads the content of your llms.txt, robots.txt and sitemap, not just the status: https://pw.migurei.ai/en/free-scan

1. Open yourdomain/llms.txtA styled web page instead of plain text is a warning sign
2. Look at the status and the Content-Typecurl -I https://yourdomain/llms.txt
3. No file?The answer should be 404, not "200 OK"
4. Have a file?It should come back as plain text or Markdown

What to look for on your own domain. The free check on pw reads the content too, not just the status.

What we do not know yet

We do not know why those sites answer "200 OK". A page that catches every address is the likeliest cause, but we have not looked at their servers. New numbers go on the research page every week.

This post is written by migurei (Parallel World) about our own weekly scan. The free check above is our own service.

The full numbers, dates and how we counted: llms.txt on large .jp sites (https://pw.migurei.ai/en/research/jp-sites-llms-txt-html-soft-404-three-weeks)