llms.txt is a proposed standard for a plain Markdown file that sits at the root of your website (yourdomain.com/llms.txt) and gives AI systems a curated summary of what your site is about and which pages matter most. Think of it as a hand-written table of contents for language models: a short description of your business, followed by a list of your most important pages with one-line explanations of each. The idea, introduced by Jeremy Howard of Answer.AI in late 2024, is that AI assistants have small context windows and struggle to parse navigation menus, JavaScript, and marketing fluff, so you give them the clean version directly.
Here is the honest part most articles skip: llms.txt is a proposal, not an adopted standard. As of 2026, none of the major AI companies (OpenAI, Anthropic, Google, Perplexity) have publicly confirmed that their crawlers or assistants use the file when generating answers. Google's search team has said its systems do not use it. So the practical answer for a service business is this: creating one takes under an hour, carries no risk, and might help at the margins, but it is nowhere near the top of the list of things that actually improve your AI search visibility. Structure, clarity, and citable content on your actual pages matter far more.
What Exactly Goes in an llms.txt File?
The format defined at llmstxt.org is deliberately simple. It is a Markdown file with a specific structure: an H1 with your site or business name, a blockquote with a one-paragraph summary, optional context paragraphs, and then H2 sections containing lists of links with short descriptions.
A service business version looks something like this:
```markdown # Digiwell Marketing
Digiwell builds AI-powered marketing systems for service
businesses: lifecycle email automation, newsletter growth,
and AI search optimization.
Services
systems audit that maps where leads leak out of your funnel.
Key Resources
The full framework we use with clients. ```
The specification also describes an optional companion pattern: publishing Markdown versions of key pages (the same URL with .md appended) so a model can read clean text instead of rendered HTML. Some developer tool companies, notably Mintlify and a number of documentation platforms, generate these automatically. That is where adoption is strongest: technical documentation, where AI coding assistants genuinely fetch and use the files.
The whole thing is designed to be read at inference time, when a user asks an assistant a question and it goes looking for answers, rather than at training time. That distinction matters, because it means the file only helps if an assistant actually requests it. Which brings us to the uncomfortable question.
Does Anything Actually Read llms.txt Yet?
The evidence in 2026 is mixed and worth stating plainly. Server logs from sites that publish llms.txt files show occasional fetches from AI-associated crawlers, but no major AI company has documented the file as part of its retrieval pipeline. Google representatives have said directly that Google's systems do not use llms.txt. Anthropic and OpenAI publish llms.txt files for their own documentation sites, which signals sympathy for the idea, but neither has confirmed their assistants consume the file from third-party sites when answering user questions.
Compare that with robots.txt, which took years to become universal but had a forcing function: crawlers that ignored it got blocked and generated complaints. llms.txt has no such enforcement mechanism. It is a courtesy offering, and adoption depends entirely on AI companies deciding it is useful. Some in the SEO community argue it will become standard infrastructure; others point out that models are getting better at parsing regular HTML, which weakens the original rationale.
Our read: treat llms.txt the way you would have treated schema markup in 2012. Early, unproven, cheap to implement, and potentially valuable if adoption arrives. What you should not do is treat it as a substitute for the work that demonstrably matters, which is making your business easier for AI to understand at the page level: clear answers, consistent entity information, and content structured so a model can quote it.
An llms.txt file will not fix a site that AI cannot parse in the first place. Get a free audit of how your site actually reads to AI search engines.
How to Create an llms.txt File in Under an Hour
If you decide the small bet is worth it, here is the build order we use. The entire process takes 30 to 60 minutes for a typical service business site.
Step 1: Write the summary blockquote (10 minutes). One paragraph, under 60 words, stating who you are, who you serve, and what you do. Write it the way you would answer "what does your company do" at a dinner party, not the way your homepage hero reads. This is the highest-leverage part of the file because it is the first thing any parser encounters.
Step 2: Choose 10 to 20 pages, not 200 (15 minutes). The file is a curation exercise. Include your core service pages, your best proof content (case studies), your pillar guides, and your contact or booking page. Leave out tag archives, thin posts, and legal pages. If everything is important, nothing is.
Step 3: Write one-line descriptions for each link (20 minutes). Each description should say what a reader gets from the page, not repeat the title. "The exact five-email sequence we build after every discovery call" beats "our blog post about follow-up emails."
Step 4: Save it as plain Markdown at your domain root (10 minutes). The file must live at /llms.txt, served as plain text. On most platforms this means dropping it in your public or static folder. Verify it loads in a browser.
Step 5: Keep it current. Add a line to your content publishing checklist: when a major page ships or a service changes, update llms.txt. A stale file that points to dead pages is worse than no file.
What Matters More Than llms.txt for AI Visibility
This is the section to act on if you only act on one. The things that measurably influence whether ChatGPT, Perplexity, and AI Overviews cite your business are on your actual pages, and they compound while llms.txt waits for adoption.
Answer-first page structure. AI engines extract passages, not pages. Pages that state the answer in the first two paragraphs get quoted; pages that build to a conclusion get skipped. The same principle applies to how you structure service pages for AI search: headings that match real questions, one idea per section, specifics a model can lift.
Consistent entity information. Your business name, location, services, and positioning should read identically across your site, your Google Business Profile, LinkedIn, and directories. Models triangulate; contradictions lower confidence.
Schema markup that is actually adopted. Organization, Service, FAQPage, and Article schema are documented, supported surfaces. Unlike llms.txt, structured data has confirmed consumers today.
Crawlability for AI user agents. Check your robots.txt and firewall rules. Plenty of sites block GPTBot, ClaudeBot, or PerplexityBot without realizing it, usually via a CDN security setting. An llms.txt file is pointless if the assistant cannot reach your site at all.
The gap between publishing an llms.txt file and doing this structural work is the gap between hoping AI notices you and giving it no choice.
Frequently Asked Questions
Is llms.txt the same as robots.txt?
No, they solve opposite problems. robots.txt tells crawlers what they may not access, and it is a long-established standard that crawlers actually honour. llms.txt tells AI systems what they should prioritize, and it remains a proposal without confirmed adoption by major AI companies. The two files can coexist: robots.txt controls access, llms.txt offers curation. Publishing an llms.txt file does not override or replace any directive in your robots.txt.
Will llms.txt improve my Google rankings?
No. Google has stated that its search systems, including AI Overviews, do not use llms.txt. The file has no effect on crawling, indexing, or ranking in Google Search. If your goal is visibility in AI Overviews, invest in clear page structure, question-matched headings, supported schema markup, and content quality. Treat llms.txt as a separate, speculative bet on future assistant behaviour, not as an SEO tactic with measurable return today.
How is llms.txt different from a sitemap?
A sitemap.xml is an exhaustive machine-readable list of every indexable URL, built for search engine crawlers. llms.txt is the opposite: a short, human-curated, annotated selection of your most important pages with descriptions, written in Markdown for language models. A sitemap says "here is everything." llms.txt says "here is what matters and why." Keep your sitemap regardless; it is required infrastructure. llms.txt is optional curation layered on top.
Does a small service business actually need llms.txt in 2026?
Need is too strong a word. No AI platform requires it, and none has confirmed using it for answering questions. Our recommendation: if you have an hour and your page-level fundamentals are already solid, publish one. The cost is trivial, the downside is zero, and you are positioned if adoption arrives. If your site still lacks clear answers, proof content, and schema, fix those first because they influence AI citations today.
Can llms.txt hurt my site if I get it wrong?
The realistic risks are small but real. A stale file pointing at deleted pages sends low-quality signals to anything that does read it. A file that contradicts your actual site content (different service descriptions, old pricing) could confuse a model that consults both. And curating carelessly wastes the one advantage the file offers. Set a quarterly reminder to review it, keep it under 20 links, and make it match your live site.
Read Next
Find Out How Your Site Reads to AI
The free audit includes an AI readability review: whether AI crawlers can reach your pages, whether your structure gives them anything worth quoting, and where competitors are getting cited instead of you. llms.txt is one line item on that checklist. The other twenty are where the visibility actually comes from.