TL;DR
- llms.txt is a plain-text file at your domain root that gives AI agents a structured summary of your business.
- Over 844,000 sites including Stripe, Cloudflare, and Vercel have already adopted it.
- Google merged an llms.txt check into Chrome's Lighthouse in April 2026, as part of its new Agentic Browsing audits.
- It reduces agent parsing cost by up to 99 percent compared to crawling your full website, and agents process structured sites first.
A brief for machines, not a page for humans
llms.txt is a plain-text markdown file hosted at yoursite.com/llms.txt. It does not appear on your website and human visitors never see it. Its purpose is to give AI agents - ChatGPT, Claude, Gemini, Perplexity, and the autonomous agents coming next - a structured summary of who you are, what you do, and how to work with you.
Think of robots.txt and llms.txt as a pair. robots.txt (1994) tells crawlers what not to access. llms.txt (2024) tells agents what to understand. One is an exclusion protocol. The other is a curation protocol.
Why agents need a dedicated file
Your website was designed for humans: images, animations, complex navigation, marketing copy. When an AI agent tries to read it, it fetches the raw HTML, strips out everything visual, and is left with roughly 18 percent of the original content. That extracted text is often noisy: nav labels mixed with footer disclaimers mixed with cookie banners mixed with your actual business facts.
An llms.txt file eliminates this extraction problem entirely. It is already clean text, already structured, already focused on the facts agents need. A typical AI Website Profile is 1,500 to 2,500 tokens, compared to 10,000 or more tokens for a multi-page crawl. The cost difference matters when agents are comparing ten or twenty vendors in a single research session.
But the cost gap does more than save compute. When an agent gets back 10 to 15 search results and has to decide which ones to actually read, it processes the cheapest, clearest sites first. Sites with llms.txt get read. Sites without it get skipped. The agent could read them, but it does not need to once it has enough signal from the structured alternatives. Good SEO gets you on the candidate list. llms.txt gets you read once you are on it.
The llms.txt specification: what the file must contain
The specification, published by Jeremy Howard at llmstxt.org, defines a Markdown file with sections in a fixed order. First, an H1 with the name of the site or business: this is the only required section. Next, a blockquote with a short summary containing the key facts needed to understand the rest of the file. Then any number of paragraphs or lists with more detail, but no headings. Finally, any number of H2 sections, each holding a list of Markdown links in the form [name](url), optionally followed by a colon and a short note.
An H2 section named Optional has a special meaning: it holds secondary links an agent can skip when it needs a shorter context. Everything else is plain Markdown. No HTML, no JavaScript, no images, and a structure precise enough that tools can parse it with ordinary code, not just language models.
The simplicity is the point. Any LLM can interpret this format natively. There is no SDK to install, no API to call, no vendor lock-in. You host a text file on your own domain and agents can find it at a predictable URL.
What changed in llms.txt v2 (August 2026)
On August 10, 2026, the specification was revised to v2 based on two years of adoption. The biggest change is scope: a file can now sit at the site root or at any path, and it covers the pages under that path. A /docs/llms.txt describes only the documentation, and when several files apply, agents should use the most specific one.
v2 also recommends publishing clean Markdown versions of important pages at the same URL with .md added, and linking pages to them with standard link relations: rel="alternate" type="text/markdown" for the Markdown version and rel="describedby" for the llms.txt file that covers the page. These can be HTML link elements or an HTTP Link header set at the server or CDN.
The specification also settles the /.well-known/ question: it considered and rejected that location, because well-known URIs exist only at the origin root and cannot describe a path.
Who is already using it
Over 844,000 websites have implemented llms.txt, including Stripe, Cloudflare, Vercel, Anthropic, and hundreds of mid-market businesses. Adoption spans SaaS documentation, e-commerce product catalogs, professional services, and local businesses. The format is gaining traction because it costs nothing to host, takes minutes to deploy, and provides immediate value to any AI system that encounters it.
The AI labs publish their own: OpenAI, Anthropic, and Google's Gemini all serve llms.txt files for their developer documentation. Documentation platforms Mintlify and GitBook generate one for every site they host, Wix generates one for every Wix site, and Yoast and AIOSEO build generators into their WordPress plugins. The standard is moving from early-adopter to mainstream infrastructure.
The clearest signal yet came in April 2026, when Google merged an llms.txt check into Chrome's Lighthouse, the same auditing tool that normalized HTTPS, mobile-first design, and Core Web Vitals. It sits in a new, experimental Agentic Browsing category. Today a missing file is marked Not Applicable because the file is still optional; a file that exists must load within 2 seconds and contain a Markdown H1 and at least one link. That is the same arc HTTPS followed: optional, then measured, then expected.
Why you should not DIY it
Creating a text file is easy. Structuring it so AI agents extract the right facts for your specific industry is not. A law firm's AI profile needs completely different data than a restaurant's or a SaaS company's. Platinum.ai uses 150 industry-specific blueprints to ensure your profile highlights the exact criteria agents look for when evaluating your type of business. We scan your existing site, resolve gaps, and deliver a production-ready file within 24 hours.
Frequently asked questions
Is llms.txt an official web standard?
Not yet a formal W3C or IETF standard. llms.txt is a community specification published at llmstxt.org, and it now has formal tooling recognition: in April 2026 Google added an llms.txt check to Chrome Lighthouse's experimental Agentic Browsing category. Over 844,000 sites have adopted it and Yoast has built support into their WordPress plugin. The trajectory matches robots.txt, which was a community convention for decades before it became an IETF standard in 2022.
Does llms.txt have to be at the root of my domain?
No. Since the v2 specification (August 2026), an llms.txt file can sit at the root or at any path, such as /docs/llms.txt, and covers the pages under that path. For a business website, the root location /llms.txt is still the right default: it describes the whole site and is where tools like Chrome Lighthouse look.
What does an llms.txt file need to include?
Only an H1 with your site or business name is strictly required. A useful file adds a blockquote summary of key facts and H2 sections listing links to your most important pages, each with a short description. Chrome Lighthouse also expects at least one Markdown link and 50 characters of content.
Does it replace my sitemap or structured data?
No. Sitemaps help crawlers discover pages. Schema markup helps search engines categorize page-level content. llms.txt gives AI agents a holistic understanding of your entire business in one lightweight file. They serve different purposes and work best together.
Will adding llms.txt hurt my existing SEO?
Not at all. It is a separate text file that sits alongside your website. It does not change your HTML, your page speed, or your search rankings. It only adds a new layer for AI agents to read.
Get your AI Website Profile built by experts
Run a free Site Scan to see your current AI readiness score. Then let us build your llms.txt using the right industry blueprint.