How to Create a /llms.txt for Your Site: The Complete Guide to Make Your Content Readable by AI Agents

Published on — Updated on — By

There is a simple text file that can change how AI agents understand your site. It’s under 50 lines, requires no advanced programming, and can be created in less than 30 minutes. It’s called `llms.txt` — and in this guide you’ll learn exactly what it is, why it matters, and how to create one for any type of site. --- ## What `/llms.txt` Is and Why It Exists `llms.txt` is a Markdown-formatted text file, hosted at the root of your domain (at `seusite.com.br/llms.txt`), that instructs artificial intelligence agents about what your site is, which pages are most relevant, and how to interpret your content. The proposal was created by Jeremy Howard, founder of fast.ai, and published in September 2024. The motivation is straightforward: language models have limited context windows and cannot process an entire website at once. When an AI agent needs to understand your business to answer a question or recommend a product, it needs a reliable, structured, concise source of information. `llms.txt` is exactly that. > "While sites serve both human readers and LLMs, the latter benefit from more concise and specialized information gathered in a single accessible place." > — Jeremy Howard, llmstxt.org Think of it as `robots.txt` — but instead of telling crawlers what not to index, `llms.txt` tells AI agents what is most important to understand about your site. --- ## The Difference Between `llms.txt`, `robots.txt` and `sitemap.xml` The three files coexist and have complementary functions. Understanding the difference avoids confusion when implementing: | File | For whom | What it does | Format | |---|---|---|---| | `robots.txt` | Search crawlers | Controls bot access to pages | Plain text with directives | | `sitemap.xml` | Search crawlers | Lists all site URLs | Structured XML | | `llms.txt` | AI agents and LLMs | Explains the site and points to relevant resources | Markdown readable by humans and machines | `llms.txt` does not replace either of the other two — it complements them by adding a semantic layer the earlier formats cannot express. --- ## Who Is Already Using It Adoption is still low across the general web: analyses of about 300,000 domains found an adoption rate of only 10.13%, and only 3 of the world’s top 1,000 sites have the file. But among companies that deeply understand AI, adoption is significant and growing. Companies like **Cloudflare**, **Anthropic**, **Vercel**, **ElevenLabs** and **Supabase** have already implemented `llms.txt`. Cloudflare goes further and maintains product-specific files. Vercel has a `llms-full.txt` so extensive it was described as "a 400,000-word novel" — the entire documentation consolidated into a single file for AI consumption. The standard is especially popular among developer tools companies, SaaS, and technical documentation — precisely the sectors where the accuracy of information consumed by AI directly impacts end-user experience. --- ## What the Science Says: Does `llms.txt` Actually Increase AI Citations? This is the most honest question to ask before investing time in implementation. The answer, based on data available up to May 2026, is: not directly — but that doesn’t mean you shouldn’t create one. The most rigorous study available, conducted by SE Ranking with about 300,000 domains and published in the Search Engine Journal, reached a direct conclusion: `llms.txt` does not appear to directly impact the frequency of citations by AI in ChatGPT, Perplexity and Google AI Mode. The study controlled for domain authority and content quality, so the result is not simply "bigger sites are cited more." Three important reasons to implement it anyway: First: None of the major platforms has publicly confirmed that they *ignore* `llms.txt`. Absence of evidence of positive impact is not evidence of zero impact — the ecosystem is still forming. Second: The file has immediate value for agents that operate in research mode (like Perplexity Deep Research and ChatGPT with active browsing). When an agent visits your site to collect information, `llms.txt` is the first thing it reads — and it determines which pages it will prioritize. Third: The implementation cost is minimal. A well-structured text file, created in 30 minutes, that can have positive impact as the ecosystem matures — the cost-benefit ratio is favorable. --- ## The Official Structure of `llms.txt` The format is defined by the specification at [llmstxt.org](https://llmstxt.org) and uses plain Markdown. The structure follows this order: **1. Required H1** — The project or site name. It’s the only truly mandatory section. **2. Optional blockquote** — A short summary with the most important information to understand the rest of the file. **3. Optional text sections** — Paragraphs or lists with detailed information about the project. They cannot contain additional headings. **4. H2 sections with lists of links** — Each H2 section delimits a "file list" with links to where more information is available. Each item is a Markdown link with an optional note after the colon. **5. Special "Optional" section** — If present, indicates that the links contained there may be ignored when context needs to be reduced. Use for secondary information. The base template: ``` # Site or Project Name > Short, objective description of what the site is, who it's for > and what its main differentiator is. Additional relevant information an AI agent needs to correctly interpret the site's content. ## Main Section - [Page Title](https://url.com/page): Description of what's on this page ## Another Section - [Title](https://url.com/other): Description ## Optional - [Secondary Page](https://url.com/secondary): May be ignored if the context is short ``` --- ## Step-by-Step Guide: Creating `llms.txt` for Each Type of Site ### For E-commerce Sites and Online Stores Focus on products, categories, purchase policies and the information an AI agent would need to recommend or buy a product on behalf of the user. ``` # Vetfleur — Natural Aromatherapy for Pets > Store specializing in aromatherapy and natural well-being > products for dogs and cats. All products are free of artificial > ingredients. Delivery across Brazil with tracking. Founded in 2019, Vetfleur develops natural formulas based on aromatherapy to promote pet well-being. Our products are veterinarian-tested and contain no parabens, sulfates or artificial colorants. ## Featured Products - [Natural Bioprotector for Pets](/produtos/bioprotetor): Natural sunscreen for dogs and cats, SPF 30, water-resistant - [Lavender Relaxing Oil](/produtos/oleo-relaxante): Aromatic blend for anxiety and canine agitation, 30ml - [Complete Well-Being Kit](/produtos/kit-bem-estar): Combo with 3 routine care products ## Categories - [Sun Protection](/categorias/protecao-solar): All sunscreens for pets - [Aromatherapy](/categorias/aromaterapia): Essential oils and diffusers for pets - [Natural Hygiene](/categorias/higiene): Natural shampoos and conditioners ## Policies - [Shipping and Delivery](/frete): Free shipping above R$ 150, delivery time 3–7 business days - [Exchanges and Returns](/devolucoes): 30-day policy for exchange or full refund - [Payment](/pagamento): Card, Pix, boleto and installment up to 12x interest-free ## Optional - [Blog](/blog): Articles on natural care and well-being for pets - [About Us](/sobre): History and values of Vetfleur - [Frequently Asked Questions](/faq): Common questions about products and orders ``` ### For Service Sites and Agencies Focus on offered services, differentiators, work process and how to get in touch. ``` # SEO 5.0 — Strategic Architecture for AI > SEO platform specialized in optimization for search engines > and generative AI systems (GEO). Automated diagnostics with the > IAS™ Score, monthly and annual plans for Brazilian companies. SEO 5.0 is Brazil's first platform focused on GEO (Generative Engine Optimization) — the practice of structuring content to be found, cited and recommended by systems such as ChatGPT, Perplexity, Gemini and Claude. ## Main Services - [IAS™ Diagnostic](/diagnostico): Free analysis of 10 visibility dimensions in AI, result in 2 minutes - [Starter Plan](/planos): R$ 129/month — ideal for freelancers and small businesses - [Professional Plan](/planos): R$ 297/month — for agencies and growing companies - [Enterprise Plan](/planos): Contact for pricing — for large operations and multiple domains ## Resources and Tools - [Blog & Insights](/blog): Technical articles on SEO 5.0, GEO and AI visibility - [Free Tools](/ferramentas): Schema analyzers, llms.txt checkers and more - [Success Cases](/cases): Real client results with SEO 5.0 ## Contact - [Talk to an Expert](/contato): Contact form and consultancy scheduling - [Support](/suporte): Help center and ticketing system ## Optional - [About SEO 5.0](/sobre): History, team and methodology - [Partners](/parceiros): Agencies and freelancer partners ``` ### For Blogs and Content Portals Focus on content categories, the most important articles and how content is organized. ``` # Sports Nutrition Blog — NutriPerformance > Content portal specialized in nutrition for amateur and > professional athletes. Evidence-based articles, > reviewed by sports nutritionists registered with the CRN. We publish weekly articles about supplementation, nutritional periodization, athlete recipes and analyses of recent research. All content is reviewed by at least one sports nutritionist. ## Main Categories - [Supplementation](/suplementacao): Complete guides on protein, creatine, caffeine and more - [Nutrition for Running](/corrida): Nutritional strategies for road and trail runners - [Recipes for Athletes](/receitas): Over 200 recipes with calculated macros - [Research and Science](/ciencia): Analyses of recent studies translated into practice ## Most Accessed Articles - [Complete Creatine Guide](/suplementacao/creatina-guia): Dosage, timing, types and evidence - [Carbohydrates Pre-Workout](/corrida/carboidratos-pre-treino): What to eat and when - [Protein: How Much Do You Really Need](/suplementacao/proteina-quantidade): Based on meta-analyses ## Optional - [About the Authors](/autores): Profiles of the nutritionists who review the content - [Editorial Methodology](/metodologia): How we select and review articles ``` --- ## The `llms-full.txt` File: When and Why to Use It In addition to the standard `llms.txt` (which works as an index), the specification foresees a complementary file called `llms-full.txt`. The difference is fundamental: - **`llms.txt`** — An index with links to where information resides. Lightweight, quick to process, ideal for most sites. - **`llms-full.txt`** — The complete content consolidated in a single file. Useful for technical documentation where the agent needs full context without making multiple requests. For most business sites, the standard `llms.txt` is sufficient. `llms-full.txt` makes more sense for API documentation, software libraries or extensive knowledge bases — exactly the use cases of companies like Vercel and Anthropic. --- ## Where to Host and How to Verify The file should be accessible at `https://yourdomain.com.br/llms.txt`. No special server configuration is necessary — it’s a static text file like any other. To verify it’s working, access the URL directly in a browser. The file should display as plain text, without surrounding HTML. Validation tools available: - [llmstxt.org](https://llmstxt.org) — Official specification with validator - [llms-text.com/generator](https://www.llms-text.com/) — Automatic generator based on URL - [directory.llmstxt.cloud](https://directory.llmstxt.cloud) — Directory of real examples for reference How to add on different platforms: On WordPress sites, create the `llms.txt` file in the root folder via FTP or your hosting file manager. In Next.js or React sites, place the file in the `public/` folder. On Shopify, use a static theme file or a custom files app. --- ## Best Practices and Common Mistakes Be specific, not generic. A `llms.txt` that says "we are a digital marketing company" doesn’t help any AI agent. Include what you do, for whom, where and how — with details that differentiate your business. Prioritize pages that convert. `llms.txt` doesn’t need to list every URL. List the pages an AI agent would need to understand your business and guide a user to purchase or contact. Keep it updated. If you launch new products, change prices or create new sections, update `llms.txt`. An outdated file can cause an AI agent to recommend incorrect information about your business. Use `## Optional` wisely. Put only truly secondary content there. If an AI agent needs to reduce context, it will ignore that section — so don’t place critical business information there. Do not duplicate `robots.txt`. `llms.txt` is not the place to block bot access. Use `robots.txt` for that. `llms.txt` is for guidance, not restriction. --- ## Implementation Checklist Before publishing your `llms.txt`, check: | Item | Checked | |---|---| | H1 with the site name present | ✓ | | Blockquote with a clear and objective summary | ✓ | | H2 sections with lists of relevant links | ✓ | | Each link has a descriptive note after the colon | ✓ | | `## Optional` used only for secondary content | ✓ | | File hosted at `/llms.txt` at the domain root | ✓ | | URL accessible directly in the browser | ✓ | | Price, availability and contact information up to date | ✓ | --- ## `llms.txt` in the Context of SEO 5.0 `llms.txt` is one layer of the **GEO (Generative Engine Optimization)** strategy — the practice of structuring a digital presence to be understood, indexed and cited by generative AI systems. It does not replace Schema.org structured data, rich semantic content or domain authority — but it complements them by adding direct readability for agents. In the context of the IAS™ Score, the presence of a well-structured `llms.txt` contributes to the **AI Interaction** dimension — one of the 10 dimensions evaluated in the diagnostic. Sites that already implement Schema.org, rich semantic content and now add `llms.txt` are building a digital presence that speaks the language of the systems increasingly mediating product and service discovery. The window of opportunity is open: only 10% of sites have the file. Those who implement it now occupy space most competitors haven’t yet realized exists. --- *Want to know which dimensions your site is losing visibility to AI agents? Get the free IAS™ diagnostic — result in 2 minutes, no signup required.*