Somewhere on your website, tucked into the root right next to robots.txt, a growing number of SEO agencies are quietly telling you to add a file called llms.txt. It is pitched as the next-generation robots.txt — the way you talk to ChatGPT, Claude, and Gemini, the modern requirement for showing up in AI search. The question a Northern Nevada plumber, flooring installer, or HVAC company should ask in 2026 is a simpler one: does the file actually do anything?
We publish an llms.txt on digitalhorizon.dev. So we asked the same question. The honest answer is more interesting than either side of the argument sounds.
What llms.txt actually is
The llms.txt proposal, introduced by Answer.AI's Jeremy Howard in September 2024, is a plain-text file that lives at yoursite.com/llms.txt. Instead of telling crawlers what to block — that job belongs to robots.txt — it hands language models a curated summary of your site. The pages worth reading, in the order a machine should read them, with brief context. Think of it less as an access rule and more as an executive summary you leave at the front door for any large language model that happens to walk up.
There is a companion file, llms-full.txt, which some sites use to include the actual page contents in one giant flat markdown document. Both are unofficial. Neither has been ratified by a standards body. No search engine or AI company is required to read either one.
The 2026 data
This is where the pitch collapses under actual measurement. In May 2026, Ahrefs analyzed 137,000 sites in its web analytics dataset. Only 3% of the llms.txt files on those sites received any AI bot requests at all. Ninety-seven percent were sitting there getting fetched by exactly nobody. SE Ranking's separate survey of nearly 300,000 domains found roughly 10% adoption. Rankability's June 2026 crawl of the world's top 1,000 websites found 8.7% publishing one.
The adoption growth is real — nearly 9x year over year. The actual usage is not. Google's Search Advocate Gary Illyes has said publicly that nobody in AI is using llms.txt. Google Search does not consume it. OpenAI, Anthropic, Meta, and Mistral have not committed to reading it in their production systems either.
Cloudflare's July 2026 crawl-traffic report added one wrinkle worth reading carefully. Plain-text fetches by AI bots doubled year over year — from 3% of AI traffic in July 2025 to nearly 6% in July 2026. Some of that is llms.txt, some is raw markdown, some is robots.txt. Adoption is growing, and a handful of secondary crawlers are checking. The volume is still small, and the ranking-critical bots — the ones that actually feed AI Overviews, ChatGPT Search, and Perplexity — largely ignore it.
Where the file does work
There are three cases where an llms.txt file earns its keep in 2026, and none of them are helping your service business rank in Google.
- AI coding assistants. Cursor, Claude Code, GitHub Copilot, and Windsurf all fetch llms.txt when a developer points them at a URL. If you publish an API, a product doc site, or a plugin developers integrate, an llms.txt tells the coding agent which docs to load into context. For a plumbing company website, that use case is thin.
- In-product AI assistants. SaaS products that build a 'chat with our docs' feature increasingly consume llms.txt as the source of truth. More relevant for a software company than a service business.
- A signal that you are paying attention. Not a ranking factor. But a well-maintained llms.txt tells the humans checking your site — competitors, journalists, prospects — that you understand the modern web. That signal has real value in a B2B context. It has less value if your customer is looking for someone to install carpet.
What Northern Nevada service businesses should actually do
For a service business in Carson City, Reno, or Sparks, the honest 2026 answer looks like this:
- Adding an llms.txt is cheap and low-risk. A ten-line file listing your service pages, your service areas, and your contact information takes fifteen minutes. It cannot hurt. It probably will not help. We would not deprioritize a single other project for it.
What actually moves the needle in AI search is different. E-E-A-T signals, a maintained Google Business Profile, structured data, citations that match everywhere, and a growing library of content clusters — those are what AI answer engines actually consume when they decide who to name for 'flooring near Carson City.'
- Do not pay an agency $500 to install an llms.txt. If someone pitches you an 'AI-readiness package' and llms.txt is on the deliverable list, ask what else is on it. If the answer is structured data, image alt text, and citation cleanup, you are paying for real work with a modern label. If the answer is 'we'll write your llms.txt file,' you are paying for a fifteen-minute task at a five-hundred-dollar rate.
Digital Horizon's own take
We publish an llms.txt on digitalhorizon.dev, and an llms-full.txt as well. Two reasons. First, we build websites for other people, and running the experiments ourselves is the only way we can tell clients whether they are worth doing. Second, we would rather explain to a prospect why we have one than explain why we don't. Neither of those reasons is 'it drives leads.' It doesn't. Nothing we track suggests otherwise.
The work that does drive leads is the same work we cover across the rest of this journal — the answer-engine-optimization playbook for showing up in AI answers, the AI Mode default tab changes that reshaped local SERPs this month, and the boring, patient signals — reviews, photos, page speed — that quietly decide which local business gets named when a homeowner in Minden asks ChatGPT for a floor installer.
The bottom line
llms.txt is the classic 2026 SEO artifact: real enough to justify a blog post, small enough to add in an afternoon, thin enough that it should not move a single dollar in your marketing budget. Add one if you want. Skip it if you don't. Neither decision changes whether the next AI answer engine names your business.
If you'd like a five-minute audit of what actually is moving the needle on your AI visibility — the specific signals that show up when a real prospect types your service into a real chatbot — we're a short conversation away. We'll pull the transcripts, tell you what's being cited, and hand you the fixes ranked by which ones move real jobs.