Pay per approved lead. First 3 leads free.

See how it works

Published in ai seo

llms.txt Study: We Checked 127 Local Business Files, and 1 in 9 Sites Turned Away Our AI-Bot Test

llms.txt study: we checked 127 local business files. 63% have a title, summary and links, and 11% turned away our AI-bot test.

llms.txt Study: We Checked 127 Local Business Files, and 1 in 9 Sites Turned Away Our AI-Bot Test
By Mohammad12 min read
In this article (13 sections)

Quick answer: RankMeHi re-checked every llms.txt file found on 151 of 489 live Orange County local business websites in November 2026. In our study, 127 were real files and 88.2% start with the one part the format requires, a title, but only 63% have a title, a summary and links. 11% refused or challenged us when we asked for the file with an AI crawler's name. No AI engine has said it uses the file, so treat it as a cheap extra.

A business owner opens their site's llms.txt for the first time and finds a page of links to old services and a phone number that has changed. They didn't write it. A plugin did, one update ago, and nobody looked at it since.

That is the file most local businesses have, if they have one. 151 of the 489 live sites in our 662-business Orange County website study answered at /llms.txt in September. RankMeHi is a web design and local SEO company in Costa Mesa, California, and we wanted to know what those files actually contain, who wrote them, and whether the sites behind them welcome the AI bots they're written for. For 1 site in 9, the file says come in and the door says not you.

Not sure what your llms.txt says? We'll read it, test your site with the AI crawlers' names and tell you what to fix, even if you never hire us. Get a free website check

How we checked the llms.txt files on 151 local business sites

Short answer: we asked each site for its /llms.txt file again in November 2026, read what came back, followed up to 25 links in each file, and repeated the request with three AI crawlers' user-agent names.

The sample is the same OpenStreetMap list of Orange County, California businesses behind our earlier study, with chains removed. Our September crawl counted 489 live sites and flagged 151 with an llms.txt file (30.9%). We went back to those 151 in November. Here is what came back, with every denominator in one place:

What we found in November 2026SitesShare of 151
A real text file at /llms.txt12784.1%
The address now returns a web page, not a file10.7%
File gone (HTTP 404 or 410)106.6%
Blocked our request (HTTP 401, 403 or 429)10.7%
Timeout or connection error (not counted either way)127.9%

Source: RankMeHi llms.txt audit, November 2026. All shares in this article are of the 127 real files unless the table says otherwise.

Three limits you should keep in mind:

  • It's a sample, not the market. One county, listed sites only. Don't read these shares as national numbers.
  • The bot test is a name, not a crawler. We sent a request from one address with the user-agent names GPTBot (OpenAI), ClaudeBot (Anthropic) and PerplexityBot (Perplexity). Real crawlers come from known address ranges, and some security tools treat them differently. A refusal means our test was refused, not that a real crawler would be.
  • Broken means broken from here. We counted HTTP 404, 410 and no connection. Answers like 403 or 429 are counted separately as blocked, because many sites refuse automated link checks.

What llms.txt is, and what Google and the studies say about it

Short answer: llms.txt is a plain markdown file at the root of a site that summarizes the site and lists its key pages for language models; no major AI company has confirmed it uses the file.

The llmstxt.org proposal from September 2024 describes it. The only required part is a title, an H1. After that come an optional summary in a quote block, optional text, and sections with lists of links.

Google has been blunt about it. John Mueller said in 2025 that no AI system currently uses llms.txt, and Google's search team has said it does not endorse the file. SE Ranking looked at nearly 300,000 domains and found no clear link between having the file and being cited in AI answers; about 10% of those domains had one. Our local sample shows 30.9%, but the samples differ, so don't compare the two numbers directly.

That gap sets up our question. If the file doesn't change what AI engines cite, why do so many local sites have it, and is it any good?

What is inside the llms.txt files local businesses publish

Short answer: most start with a title, and about 6 in 10 are complete by our measure, with a title, a summary and links.

What the file containsFilesShare of 127
A title (H1), the only part the format requires11288.2%
A short summary in a quote block8970.1%
At least one list of links11590.6%
Section headings (H2)11993.7%
Title, summary and links together8063%
Under 200 bytes (a stub)64.7%

Source: RankMeHi llms.txt audit, November 2026. The format requires only the title; "complete" is our stricter measure.

Bar chart of what is inside the llms.txt files of 127 Orange County local business sites: 88.2% have a title but only 63% have a title, summary and links
A title is easy. A title, a summary and links is what makes the file useful.
Use this chart on your site

Free to use under CC BY 4.0. Please credit RankMeHi with a link to this page. Copy the code:

The median file runs 490 words and lists 10 links. 12 of the 127 have no link list at all, which makes them a title and some prose, not a guide to anything.

A file that is 63% complete isn't a failure. The format is short and forgiving, and that's the point of it. The risk is staleness: a file written once and never read again keeps pointing at whatever the site looked like that day.

Open wooden drawer of a library card catalog full of index cards, a picture of what an llms.txt file is meant to be: a short index to the site
llms.txt is an index card for the site, not the library.

Check yours in five minutes: open yoursite.com/llms.txt. Read the first three lines and the first five links. If the summary is wrong or a link is dead, a customer or a model reading it sees the same thing you do.

Who wrote the llms.txt files: plugins and website builders

Short answer: about 7 in 10 files likely came from a plugin or a website builder, not from the business owner.

We looked for plugin names in the files themselves, and we noted when the site runs on a hosted builder that offers the file. A builder match is a likely source, not proof for that file.

Where the file seems to come fromFilesShare of 127
No generator sign3628.3%
Site on Wix2620.5%
Yoast plugin text in the file2116.5%
Site on GoDaddy Builder118.7%
Site on Shopify118.7%
Site on Duda75.5%
AIOSEO plugin text in the file64.7%
Rank Math plugin text in the file53.9%
Site on Squarespace32.4%
Site on Webflow10.8%

Source: RankMeHi llms.txt audit, November 2026. "Likely automatic" is 91 of 127 files (71.7%); 36 files (28.3%) show no sign of a generator, and some of those were probably written by hand.

The pattern by platform is where it gets interesting:

Website platformFilesWith title, summary and links
WordPress4027 of 40 (67.5%)
Wix2624 of 26 (92.3%)
Custom or unknown209 of 20 (45%)
GoDaddy Builder1313 of 13 (100%)
Shopify110 of 11 (0%)
Duda70 of 7 (0%)

Source: RankMeHi llms.txt audit, November 2026. Platforms with fewer than 7 files are left out.

A file made by a tool isn't bad. It's just not a decision. The WordPress files depend on whichever SEO plugin was installed, and the quality follows that plugin's defaults. Files from Shopify and Duda sites in our sample had a title and links, but no summary quote block, so none met our "complete" measure. That's a template choice, not a business owner's choice.

Short answer: most files have working links, but 7 of 115 files with links contain at least one that is gone.

We checked up to 25 unique links per file: 1621 links in 115 files. 31 of them were broken (HTTP 404, 410 or no connection), which is 6.1% of the files that had links. Another 37 refused our request with a 401, 403 or 429, and we did not count those as broken.

The share is small, and it's also the one number in this study where the harm is easy to explain. A model or a person who follows the link finds nothing. A file the business never reads can't tell you that a service page moved.

Check yours in five minutes: click every link in your file. Fix or remove any that fail, and send the file's author a note if a plugin produces it.

Do sites publish llms.txt and turn AI bots away?

Short answer: 14 of 127 sites (11%) answered at least one of our AI-bot tests with a block page, a 403 or an error, while the same file loaded for a normal browser.

For each real file we requested it again with a bot name. Here is what happened:

User-agent name we sentFiles not servedShare of 127
GPTBot118.7%
ClaudeBot118.7%
PerplexityBot97.1%
At least one of the three1411%

Source: RankMeHi llms.txt audit, November 2026. A user-agent test from one address, not real crawler traffic.

Bar chart of the share of 127 llms.txt sites that did not serve the file to GPTBot 8.7%, ClaudeBot 8.7% and PerplexityBot 7.1%, and 11% for at least one
About 1 site in 9 turned away at least one bot name we tried.
Use this chart on your site

Free to use under CC BY 4.0. Please credit RankMeHi with a link to this page. Copy the code:

Glass shop door at dusk with a lit OPEN sign and a padlock and chain across the handles, a picture of an llms.txt file that says come in on a site that turns bots away
The file says open. The door is chained.

Our September crawl showed the same pattern for 12 of these sites, so it isn't a one-off. We don't know which of those sites chose this. A security setting, a hosting plan or a plugin may have done it for them. Our earlier AI crawlers study found that few sites block AI bots in their robots.txt file (the standard behind it is RFC 9309) and that the blocking happens at the security layer; this adds a plain version of it: the file that says "read me" sits behind a door that might say no.

If you want AI assistants to read your site, one test tells you more than the file does. Ask for your pages with those same bot names and see what comes back.

Does llms.txt help you get cited by AI? What we can and can't say

Short answer: we can say what the files contain and whether they load; we can't say whether any AI engine uses them, and nobody outside those companies can.

Our test doesn't measure citations. It checks files, links and access. Google's John Mueller has said no AI system uses it, and the largest public study we found saw no clear effect. So our view is simple: llms.txt is a cheap, optional extra, and a file with wrong facts is worse than none.

We run the same checks on our own file. RankMeHi's llms.txt at rankmehi.com has a title, a summary, 7 sections and 30 links; none of the 25 we checked were broken, and all three of the three bot-name tests got the file. It was last modified on 21 November 2026. That's one site and a point in time, not a guarantee. If it drifts, you can check us.

What to do about llms.txt: a five-step order

Short answer: fix access first, then facts, then links, then structure, and only then decide whether you want a file at all.

The five steps, in order

  1. Request your site with an AI bot's name (GPTBot, ClaudeBot, PerplexityBot) and note what comes back. *Out of order:* a perfect file behind a block page reaches nobody.
  2. Read your file as a customer would. Fix the summary, the phone number and every service that no longer exists. *Out of order:* fixing links first sends readers to the wrong place faster.
  3. Click every link and remove or repair the dead ones. *Out of order:* links to pages that will later move have to be fixed twice.
  4. Add the missing parts a good file has: a one-paragraph summary in a quote block, and sections with links to your main service pages. *Out of order:* structure on top of old facts hides errors under a tidy layout.
  5. Decide who owns it. If a plugin writes it, put a reminder to read it each quarter; if you write it, keep it short. *Out of order:* nothing breaks, but without an owner it goes stale again.

Why the order matters, and what it costs

Do step 2 before step 1 and you've repainted a sign on a door that stays locked; skip step 2 and you've unlocked the door onto the wrong address. In our experience, steps 2 to 5 take under an hour for a small site. Step 1 takes a minute with a tool, and longer if your host or security plugin needs a setting changed.

For the wider picture of what AI search looks at, our answer engine optimization guide covers the pages and facts that matter more than any single file, and our AI search optimization service covers the work behind it.

When llms.txt isn't worth your time

Short answer: if your site is missing basic pages, loads slowly, or hides its phone number, fix those before you touch llms.txt.

We have measured those problems too. In our local business website statistics, 26.4% of live sites had no tap-to-call link and 24.9% had no form or booking. A customer who can't call you won't be rescued by a file for machines. If your basics are in place and you use a plugin that writes the file, read it once and move on.

What RankMeHi does about llms.txt and AI crawlers

Short answer: a free check first, then a fixed-price fix if you want one.

Our free website check includes your llms.txt file, the bot-name test above, and the basics a customer sees. If you want us to fix what we find, our AI readiness sprint starts from $390 and takes under a week, and audit and fix work starts from $490. You see the price before anything starts. Whenever we design a website or build one, we make sure the files that describe it match the pages that exist.

Key takeaways

  • RankMeHi's November 2026 audit found 127 real llms.txt files among 151 flagged live Orange County local business sites.
  • 88.2% have a title, the only required part; 63% have a title, a summary and links.
  • About 71.7% likely came from a plugin or a website builder, and 7 of 115 files with links hold a broken one.
  • 11% turned away at least one of our AI-bot tests; it was a user-agent test, not real crawler traffic.
  • No AI engine has confirmed using llms.txt; keep the file short, true and checked.

FAQ

What is llms.txt?

llms.txt is a markdown file at the root of a website, at yoursite.com/llms.txt, that summarizes the site and lists its main pages. The proposal came out in September 2024. Only a title is required; a summary and sections of links are optional. It's meant to help language models read a site.

Does llms.txt help you rank or get cited by AI?

No one has shown that it does. Google has said no AI system uses it, and a study of nearly 300,000 domains found no clear link to AI citations. In our audit of 127 local files, we measured contents, links and access, not citations. Treat it as a cheap extra.

Do I need an llms.txt file for my local business?

No. It's optional, and your customers don't see it. If your plugin or website builder already creates one, read it once for wrong facts and dead links. If you don't have one, fix your phone link, forms, speed and pages first.

How do I check my llms.txt file?

Open yoursite.com/llms.txt in a browser. Check that the title and summary are true, click each link, and make sure the address returns a text file, not a web page. Then request it with an AI crawler's user-agent name to confirm your security settings don't turn it away.

Why did our site block AI bots?

Often it's a setting nobody chose: a security plugin, a hosting plan or a bot filter. In our test, 11% of sites with a file refused at least one bot name, and we can't say who decided. Ask your host or developer which rules apply to GPTBot and ClaudeBot.

What should an llms.txt file contain?

A title with the business name, a one-paragraph summary in a quote block, and a few sections of links to your main pages, each with a short description. Keep every fact true: phone number, services, service area. Our own file has a title, a summary, 7 sections and 30 links.

Cite this study

RankMeHi, "llms.txt Study: We Checked 127 Local Business Files," November 2026. Sample from the RankMeHi 662-site Orange County local business website study; files re-checked in November 2026. More numbers from the same sites: local business website statistics. You may use the charts with a link to this page.

Find out what your files say. We'll read your llms.txt, test your site with the AI crawlers' names and send the fixes in order. Get a free website check

RankMeHi designs and builds business websites, then makes sure they are found and convert. Start with the two pages below.

Check availability

Is your trade still open in your city?

One partner per trade per city. Send your website and within one business day you will know whether your city is open, what your current site is losing, and the per-lead price for your trade.
  1. 1Share your name, email and website.
  2. 2We check whether your trade and city are still open, and audit your current site.
  3. 3You get the audit and, if your city is open, the partner terms in writing.

The audit is free either way. If the partner deal is not a fit, you still get the findings and the fixed-price quote.

Check if your city is open

Name, email and your trade. We reply within one business day with availability and the per-lead price for your trade.

Your contact details
What are you after?0 of 2 required fields done

We check whether your trade and city are still open and reply with the per-lead price.

Only for the reply. Never shared or sold.
Optional. Leave it blank if you have none yet
Optional
Optional, but it speeds up the reply
How did you find us?Optional
  • Reply within 1 business day
  • No spam, no obligation

By submitting this form, you agree that RankMeHi may contact you about your request.