llms.txt is a plain-text file that points AI tools to a site's main pages. On Nov. 1, 2026, a third of the US home-service contractor websites we checked in Google's local map results had one, and software made most of those files. In a separate graded set, AI answers named businesses with a file 40.1% of the time and the rest 41.1%. Use it as a concise index of the pages that explain your business.
Last verified: November 1, 2026.
Keep an accurate file if your platform makes one. If you're building one, keep it short and link to your most useful service, location and contact pages. The work starts on those pages: the index should point to clear, current business information.

What llms.txt does
Jeremy Howard proposed llms.txt on Sept. 3, 2024, mainly for AI tools that read a site while they answer someone's question, and it's still a proposal rather than a web standard. The file is Markdown saved as plain text at the root of a site. Its one required line is a top heading with the site's name, and everything after that is optional: a one-line summary, lists of links, and a section titled "Optional" for pages that matter less.
It isn't robots.txt, so it doesn't grant or block anything, and no crawler is obliged to open it. Google's guide to its AI features in Search, updated July 10, 2026, says the file "will neither harm nor help your site's visibility or rankings in Google Search," because Search ignores it, and that you don't need AI text files or Markdown to show up there. For what those features do read, start with how local businesses show up in Google AI search.
A short file for a home-service company looks like this (the business is made up):
# Example Plumbing & Heating
> Licensed plumbing and water heater company serving Springfield and towns within 30 miles. Emergency service 24/7.
Phone: (555) 555-0123
## Main pages
- [Emergency plumbing](https://www.example.com/emergency-plumbing): what we fix after hours
- [Water heaters](https://www.example.com/water-heaters): repair and replacement
- [Service area](https://www.example.com/service-area): every town we serve
- [Contact](https://www.example.com/contact): phone, hours and addressEvery fact in it should already appear on the pages it links to and match your Google Business Profile, because the file only points at those pages.
A third of contractor websites in Google's local map results have one
Ahrefs and SE Ranking both measured llms.txt across the web this year, but neither broke out local service businesses, so we checked 1,096 US home-service contractor websites from Cheers' scans of Google's local map results, with Cheers customers removed. On Nov. 1, 369 of them (33.7%) served a valid llms.txt, meaning a text file rather than a web page.
Another 70 (6.4%) answered /llms.txt with an ordinary web page, which a checker that only looks for a successful response will count as a file.
Cheers contractor website census, Nov. 1, 2026
A third of the contractor websites we checked serve an llms.txt file
33.7%
served a valid llms.txt file
Base: 369 of 1,096 contractor websites, Cheers customers removed
54.2%
of those files came from a site builder or SEO plugin: Wix, Duda, Yoast, Rank Math or All in One SEO
200 of 369 files, a floor
6.4%
of sites answered /llms.txt with an ordinary web page instead of a text file
70 of 1,096 sites
32.0%
of the files we read list a phone number, one more place it has to stay current
115 of 359 files read again the same morning
Cheers contractor website census, 1,096 US home-service contractor websites crawled Nov. 1, 2026, 06:57 to 07:00 UTC; files read again 10:40 to 10:57 UTC
| Finding | Value | Base |
|---|---|---|
| served a valid llms.txt file | 33.7% | 369 of 1,096 contractor websites, Cheers customers removed |
| of those files came from a site builder or SEO plugin: Wix, Duda, Yoast, Rank Math or All in One SEO | 54.2% | 200 of 369 files, a floor |
| of sites answered /llms.txt with an ordinary web page instead of a text file | 6.4% | 70 of 1,096 sites |
| of the files we read list a phone number, one more place it has to stay current | 32.0% | 115 of 359 files read again the same morning |
Files built for AI agents are rarer still. None of the sites published an MCP server card of its own, and one served an AI Catalog.
Software wrote most of these files
Wix generates and maintains the file for sites on an upgraded plan with a connected domain and search indexing turned on. Duda's June 27, 2025, product update says it writes a new one every time a site is published, and All in One SEO's feature is on by default. Those three defaults made 125 of the files (33.9%).
Another 75 (20.3%) came from Yoast or Rank Math, and both need someone to switch the feature on; Rank Math's file stays empty until someone configures it. That's 200 files from software (54.2%), and it's a floor, since custom builds and other generators leave no signature we could read. If your site has an llms.txt, there's a good chance nobody at your company decided to add it, and that's fine.
The platform a site runs on mostly decides whether it has one. Of the 976 homepages we could read, all 49 on Duda served a detected text file (100%), as did 52 of 55 on Wix (94.5%) and 32 of 43 on GoDaddy (74.4%), a rate we found no GoDaddy documentation to explain. Custom Next.js builds had one at 17 of 30 (56.7%) and WordPress at 148 of 419 (35.3%), where 98 of those files carry a Yoast, Rank Math or All in One SEO signature. Sites whose platform we couldn't identify came in at 56 of 315 (17.8%), and Squarespace at 3 of 30 (10.0%).
Who put the file there
Whether a contractor site has llms.txt mostly depends on what it's built on
Sites with a valid llms.txt
49 of 49 sites; Duda writes the file on every publish
52 of 55 sites; Wix generates and maintains the file
32 of 43 sites; we found no GoDaddy documentation explaining this
17 of 30 sites
148 of 419 sites; 98 of the 148 files carry a Yoast, Rank Math or All in One SEO signature
56 of 315 sites
3 of 30 sites
Cheers contractor website census, 976 readable homepages of 1,096 US home-service contractor websites, crawled Nov. 1, 2026
| Website platform | Sites with llms.txt (%) |
|---|---|
| Duda | 100% (49 of 49 sites; Duda writes the file on every publish) |
| Wix | 94.5% (52 of 55 sites; Wix generates and maintains the file) |
| GoDaddy | 74.4% (32 of 43 sites; we found no GoDaddy documentation explaining this) |
| Next.js (custom builds) | 56.7% (17 of 30 sites) |
| WordPress | 35.3% (148 of 419 sites; 98 of the 148 files carry a Yoast, Rank Math or All in One SEO signature) |
| Platform not identified | 17.8% (56 of 315 sites) |
| Squarespace | 10% (3 of 30 sites) |
What's inside a typical contractor llms.txt
We read the files again later that morning and could open 359. The median file is 3.3 KB, and 186 (51.8%) are mostly a list of page links. A third list a phone number (115, 32.0%), 118 (32.9%) mention reviews, ratings or stars, and 35 (9.7%) contain a dollar amount. If nobody owns the file, nobody changes those lines when the number or the price does.
Some files are just broken. Two name the business "My Site", a builder's placeholder nobody replaced, and eight have no top heading even though it's the one required line. Five contain robots.txt rules (User-agent lines with Allow or Disallow), which do nothing inside llms.txt.
A file that repeats an old phone number is one more copy of the wrong number online, and how to fix wrong business information in ChatGPT starts with the sources AI reads first.
Were businesses with llms.txt named by AI more often?
This is a different set from the census. The Cheers AI visibility grader asks ChatGPT, Gemini and Perplexity the questions a buyer would ask about a trade in a city, and we split 356 graded websites, most of them home-service companies, by whether they had a detected text file on Nov. 1.
Averaged per website, answers named the 156 sites with a file about a point less often than the 200 without. The 95% range for that gap runs from 8.5 points lower to 6.0 points higher, but that observational interval does not establish a causal effect of adding the file.
Review count could hide a real effect, because businesses with 1,000 or more Google reviews were named far more often than the rest, so we also compared sites within review bands. Weighted across the bands, the gap was 2.1 points against the file. With a file and without, the rates were 35.3% and 33.2% under 200 reviews, 34.0% and 44.1% from 200 to 999, and 58.0% and 56.0% in the top band. No band's gap is larger than chance, so the middle band doesn't show that llms.txt hurts.
The observed comparison
Graded businesses with llms.txt were named in AI answers no more often
Cheers AI visibility grader, 383 businesses on 356 readable websites, 4,592 answers from ChatGPT, Gemini and Perplexity developer APIs with web search, Mar. 17 to Oct. 30, 2026 (Pacific time); llms.txt checked Nov. 1, 2026
| Graded websites | Position | Answers naming the business (%) |
|---|---|---|
| All 356 graded websites (156 with, 200 without) | With llms.txt | 40.1% |
| All 356 graded websites (156 with, 200 without) | Without | 41.1% |
| Under 200 Google reviews (67 with, 94 without) | With llms.txt | 35.3% |
| Under 200 Google reviews (67 with, 94 without) | Without | 33.2% |
| 200 to 999 reviews (51 with, 54 without) | With llms.txt | 34% |
| 200 to 999 reviews (51 with, 54 without) | Without | 44.1% |
| 1,000+ reviews (34 with, 46 without) | With llms.txt | 58% |
| 1,000+ reviews (34 with, 46 without) | Without | 56% |
Two much larger studies point the same way. Ahrefs (June 15, 2026) found a valid file on 28% of the 137,210 domains in its web analytics, a figure it calls an upper bound, and 97% of those files got zero requests in May. SE Ranking (Oct. 15, 2026) checked nearly 300,000 domains, found llms.txt on 10.13% of them and saw no correlation between the file and how often AI cited a site.
AI bots visited our site every day and asked for llms.txt 8 times
On cheers.tech, our own site, AI bots made 25,327 requests from Oct. 18 to Oct. 31 (UTC), and no day had fewer than 930. Eight were for llms.txt: Amazonbot sent six, ClaudeBot one and ChatGPT agent one. OpenAI's search crawler made 5,618 requests and Meta's two crawlers made 11,432 between them, and not one asked for the file. llms-full.txt, the longer version we also publish, drew four more.
Our own site's logs
AI bots made 25,327 requests to our site in two weeks. 8 were for llms.txt.
8
requests for llms.txt from AI bots
Base: 25,327 AI bot requests to cheers.tech
5,618
requests from OpenAI's search crawler, none of them for llms.txt
OAI-SearchBot
11,432
requests from Meta's two crawlers, and not one asked for llms.txt
Meta-ExternalAgent and Meta-WebIndexer
4
requests for llms-full.txt, the longer version we also publish
all AI bots, same window
cheers.tech request logs, Oct. 18 00:00 to Nov. 1 00:00 UTC, 2026
| Finding | Value | Base |
|---|---|---|
| requests for llms.txt from AI bots | 8 | 25,327 AI bot requests to cheers.tech |
| requests from OpenAI's search crawler, none of them for llms.txt | 5,618 | OAI-SearchBot |
| requests from Meta's two crawlers, and not one asked for llms.txt | 11,432 | Meta-ExternalAgent and Meta-WebIndexer |
| requests for llms-full.txt, the longer version we also publish | 4 | all AI bots, same window |
Who does ask for it? Counting every visitor, our llms.txt got 305 requests in those two weeks, and automated clients our host couldn't identify sent 233 of them; SEO audit tools sent 16.
Treat this as one site, ours, with an unusual audience of marketers and SEO tools. We serve a 62.6 KB llms.txt that anyone can open, so these logs show what a real, filled-in file draws there, but they can't tell you how often bots ask for a contractor's file.
The same AI bots requested robots.txt 988 times, since that's the file where crawler permissions live. OpenAI's crawler docs say OAI-SearchBot and GPTBot take their settings from robots.txt, while ChatGPT-User visits start with a person, so robots.txt rules might not apply to them. On Oct. 30, Google's John Mueller wrote that Google aims to support a shared AI-training opt-out in robots.txt, along with other AI providers, in 2027. Contractor sites rarely connect the two: only 17 of the 947 valid robots.txt files in our census mention llms.txt at all.
Next step
Is AI recommending your business?
Find out how visible you are across ChatGPT, Gemini, Perplexity, and AI Overviews.
What to do with llms.txt this week
Type your domain plus /llms.txt into a browser. A 404 (page not found) means you don't have one. If the URL returns an ordinary web page, have your web vendor serve the intended text file or a proper 404 so the URL accurately describes what's available. If a text file comes back, read it and fix or delete any wrong business name, phone number, service area or price.
If Wix, Duda, All in One SEO, Yoast or Rank Math made the file, leave it on, and ask whoever runs your site if you're not sure which tool did. On Wix, fix wrong details on your site's pages first, because a hand edit to the file stops its automatic updates until you reset it, and the file then drifts out of date as your pages change.
Open /robots.txt next, because that's the file crawlers check for permission, and make sure the search and user-triggered AI bots you want can reach your service, location and contact pages. Which AI crawlers to allow covers each bot by name.
Put the facts an AI repeats on pages a customer can open: hours, service area, phone, license and prices, in visible text and in LocalBusiness markup. What a location page should include is the checklist, and What Is JSON-LD? shows the markup.
Include the file in the same maintenance record as the pages and profiles it describes: what changed, when it changed and who checked the result.
Running a portfolio? Add one line to the quarterly site audit ("does /llms.txt exist, and are the name and phone right?") instead of opening a project per site.
If you're the web vendor, keep whatever file the platform generates and check its name and phone against the Google Business Profile. Maintain it alongside the service and location pages and their markup.
We publish an llms.txt on cheers.tech as a concise guide to our content. Our approach to AI visibility combines accessible pages, accurate business information and useful machine-readable formats. Website optimization is how we run the page work across many locations.
Method
Census. The listings come from five weekly Cheers scans of Google's local map results, starting Sept. 28, 2026, in markets where 10 Cheers home-service customers operate. They aren't a national random sample, and a few markets supply most of the listings. We kept every website that held a top-three spot in at least half its appearances plus random draws of the rest, removed directories, social sites, platforms, government and big-box domains and Cheers customers, and set aside sites that aren't US home-service contractors, which had the file less often. That left 1,096. On Nov. 1 from 06:57 to 07:00 UTC we requested each homepage, /robots.txt, /llms.txt and /.well-known/mcp.json once with a desktop Chrome user agent. Our detection rule counted text files over 20 characters, rather than HTML pages. It did not validate compliance with the llms.txt specification. We read the first 8 KB of each file again from 10:40 to 10:57 UTC (10 sites didn't answer); "mostly links" means link bullets make up at least 80% of body lines. Platform comes from homepage code, and we left out platforms with fewer than 30 sites. Software counts are a floor. Weighting the groups back to all listings in these markets gives about 33%.
Graded comparison. The grader asks ChatGPT, Gemini and Perplexity through their developer APIs with web search on, not the consumer apps. For each business it asks four buyer questions built from its category and city ("Who are the best [trade] companies in [city]?"), each as a fresh request with no sign-in, history or personalization. Since Sept. 23, ChatGPT and Perplexity also get the business's city as the asker's approximate location, and in late September we reworded the buyer questions. From March 18 to June 8, ChatGPT answered all four buyer questions in one request for about a third of businesses, a method since retired.
The sample is 383 businesses on 356 websites, 4,592 answers from Mar. 17 to Oct. 30, 2026 (Pacific time), with Cheers customers' websites removed; it shares 6 websites with the census. Review bands held 67 sites with a file and 94 without under 200 reviews, 51 and 54 from 200 to 999, and 34 and 46 at 1,000 or more. On the 259 home-service sites alone, the rates were 34.1% with a file (118 sites) and 37.6% without (141). We checked llms.txt on Nov. 1, after the answers came in, so a site may have added or dropped its file in between. It's an observational comparison, so it can't say what adding or removing the file would do for one business, and sites with a file may differ in ways we didn't measure.
Our logs. cheers.tech request logs from Oct. 18 00:00 to Nov. 1 00:00 UTC. "AI bots" are requests our host's verified-bot classification labels as AI crawlers, assistants or agents. Counts are requests of any status, so a repeated redirect counts twice, and bots that don't identify themselves aren't counted.
Cite as: Cheers llms.txt census, Nov. 1, 2026: 1,096 US home-service contractor websites from Google's local map results, plus a comparison of AI answers for 356 graded business websites, Mar. 17 to Oct. 30, 2026 (PT). Dylan Allen-Arnegard, Cheers.
Sources
- The /llms.txt file, Jeremy Howard (llmstxt.org), Sept. 3, 2024. The original proposal and the file's format.
- Optimizing your website for generative AI features on Google Search, Google Search Central, updated July 10, 2026. Google Search ignores llms.txt.
- Understanding your site's llms.txt file, Wix Help Center, undated. Wix generates the file and when a hand edit stops updates.
- Duda is prepping your sites for the future of AI search, Duda, June 27, 2025. Duda writes a new file on every publish.
- How to Create an LLMS.txt Using All in One SEO, AIOSEO, updated Aug. 14, 2026. The feature is on by default.
- Yoast launches the first llms.txt generator in an SEO plugin, Yoast, June 2, 2025. Yoast's generator needs switching on.
- How to Use llms.txt in Rank Math SEO, Rank Math, undated. The module is off by default and empty until configured.
- We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read, Ahrefs, June 15, 2026. Web-wide adoption and request counts.
- Does LLMs.txt impact your AI visibility and citations?, SE Ranking, Oct. 15, 2026. No correlation with AI citations.
- Overview of OpenAI crawlers, OpenAI, undated. Which OpenAI bots follow robots.txt.
- Shared AI training control, John Mueller, Google Search Central Blog, Oct. 30, 2026. Google's planned robots.txt AI-training opt-out.
Sources read November 1, 2026.
Dylan Allen-Arnegård is the CEO & Co-Founder of Cheers, the done-for-you local visibility partner for established home-service companies and PE-backed portfolios.
First published December 1, 2025 by Amadeus Peterson; rewritten November 2026.
