How to implement llms.txt for your business website

What is llms.txt and why does it exist?

Every day, hundreds of millions of people ask AI assistants like ChatGPT, Perplexity, Claude, and Google Gemini for business recommendations. When those systems research your company, they face a challenge: your website was designed for humans. Navigation menus, marketing copy, legal pages, cookie banners, and JavaScript-rendered content all create noise that AI models must filter through to find the essential facts about your business.

The llms.txt specification solves this by providing AI systems with a pre-curated summary. It is a simple Markdown file placed at https://yourdomain.com/llms.txt that contains your business name, a description of what you do, your services, key facts, and links to your most important content. The format is intentionally minimal — no complex schema, no registration, no tooling required.

The numbers make the case clear:

• ChatGPT surpassed 900 million weekly active users in early 2026, many asking for direct business recommendations.
• Gartner projects a 25% decline in traditional search volume by end of 2026, replaced by AI-mediated discovery.
• 60% of Google searches now end with zero clicks — AI systems consume and summarize your content without the user ever visiting your site.
• Only 11% of domains are cited by both ChatGPT and Google AI Overviews for the same queries, indicating that AI citation remains a competitive frontier.

A well-crafted llms.txt file ensures that when an AI agent evaluates your business, it works with accurate, complete, and current information — rather than inferring details from scattered pages or outdated training data.

How llms.txt differs from robots.txt and sitemap.xml

These files are complementary. robots.txt tells crawlers where they can go, sitemap.xml tells them what exists, and llms.txt tells AI systems what it all means. A complete AI visibility strategy uses all three. Neglecting any one layer creates a gap in how AI systems perceive your business.

The llms.txt format specification

The format follows a straightforward Markdown structure with these rules:

• The file must be hosted at https://yourdomain.com/llms.txt — the root of your domain, not a subdirectory.
• It uses standard Markdown: # H1 for the business name, ## H2 for sections, - for list items.
• The first H1 heading is your business or site name.
• A brief paragraph immediately after the H1 serves as the general description.
• Subsequent ## H2 sections organize information by category (Services, Key Facts, Content, etc.).
• Links use standard Markdown format: Link text: Description
• Optionally, you can provide an llms-full.txt file with expanded content — detailed documentation, service descriptions, and in-depth material.

The specification is deliberately simple. There are no required fields beyond the H1 heading, no validation endpoints, and no formal registration process. This low barrier to entry is by design — it encourages adoption while ensuring the file remains easy for both humans and machines to read.

Step-by-step implementation guide

Step 1: Define your business identity

Start with a clear H1 heading and a one-to-three-sentence summary of what your business does, who it serves, and where it operates. This is the most critical section — it sets the context for everything that follows. AI agents often read just this portion when performing initial evaluations.

Example:
# Numinam
Numinam is a web agency based in Brussels, Belgium, specializing in high-performance websites, conversion optimization (CRO), and AI visibility (GEO / Agentic GEO) for B2B companies and SMEs across Belgium and Europe.

Be specific about your location, your specialties, and your target market. Vague descriptions like “We help businesses grow online” give AI agents nothing to work with when matching you to user queries.

Step 2: List your core services

Create a ## Services section with a clear list of what you offer. Each entry should include a link to the relevant page and a concise description. AI agents use this section to match your business against specific user needs.

Example:
## Services
- Web Design & Development: Custom, high-performance websites built with Astro…
- SEO & GEO: Search engine optimization combined with Generative Engine Optimization…
- Conversion Optimization: Data-driven CRO to increase lead generation…

Step 3: Add key business facts

Include a section with factual, structured information that AI agents commonly need: location, pricing ranges, target market, languages, and differentiators. This is what agents use when comparing you against competitors.

Example:
## Key Facts
- Location: Brussels, Belgium
- Service area: Belgium, France, Luxembourg, Europe
- Founded: 2024
- Starting price: Website projects from EUR 3,000
- Languages: French, English, Dutch

Step 4: Link to important content

Point AI systems to your most valuable content — blog posts, case studies, guides, and resources. This helps agents build a richer understanding of your expertise and authority.

Example:
## Key Content
- Agentic GEO Guide: How to prepare your website for AI agents in 2026.
- B2A2C Model: The new Business to Agent to Consumer framework.
- Website Costs: Comprehensive pricing guide for websites.

Step 5: Deploy the file

Save the file and ensure it is accessible at https://yourdomain.com/llms.txt. The deployment method depends on your platform:

• Static site generators (Astro, Next.js, Hugo, Gatsby): Place the file in your public/ or static/ directory. It will be served at the root automatically.
• WordPress: Use a plugin like “Headers & Footers” to create a custom route, or upload the file directly to your web root via FTP/SFTP.
• Shopify: Add a custom page template or use the files API to host the file at the correct URL.
• Custom server (Node.js, PHP, etc.): Add a route that serves the raw Markdown file with a text/plain or text/markdown content type.

Test by opening https://yourdomain.com/llms.txt directly in your browser. You should see plain Markdown text — no HTML rendering, no styling, no navigation.

What to include for different business types

How AI crawlers and agents use llms.txt

AI systems interact with llms.txt in several distinct ways depending on their architecture:

• Direct retrieval: AI agents with browsing capabilities (ChatGPT, Perplexity, Gemini) can fetch llms.txt directly when researching a domain. This gives them a structured summary without crawling multiple pages.
• Training data enrichment: As llms.txt files become indexed by search engines, they may be included in future LLM training corpora, improving the model’s baseline knowledge of your business.
• Agent decision-making: In the emerging B2A2C model, AI agents comparing businesses on behalf of users can use llms.txt to extract structured facts — pricing, service area, specialties — without parsing complex web pages.
• RAG pipelines: AI systems using Retrieval-Augmented Generation can pull llms.txt as a high-signal document when generating answers about your business or industry.

The key insight is that llms.txt reduces the inference burden on AI systems. Instead of requiring the model to extract facts from marketing-heavy web pages, you provide those facts directly in a clean format. This reduces hallucination risk and increases the likelihood of accurate, favorable representation.

Testing and validation

After deploying your llms.txt file, verify it works correctly:

• Browser check: Navigate to https://yourdomain.com/llms.txt and confirm the raw Markdown renders correctly. Check for encoding issues, broken links, or formatting problems.
• HTTP headers: Use curl -I https://yourdomain.com/llms.txt to verify the server returns a 200 status code and an appropriate content type (text/plain or text/markdown).
• Markdown validation: Paste your file into a Markdown previewer to confirm the structure renders as expected — headings, lists, and links should all be syntactically correct.
• AI agent testing: Ask ChatGPT or Perplexity about your business and compare their answers against the information in your llms.txt. If the AI gets key facts right that it previously got wrong, the file is having an impact.
• Link verification: Click every link in the file to confirm it resolves to a live page. Broken links in llms.txt undermine its value as a trusted source.

Common mistakes to avoid

• Being too vague. “We offer great solutions for businesses” tells an AI nothing actionable. Be specific: what services, what prices, what locations, what industries.
• Keyword stuffing. llms.txt is not an SEO page. AI models parse for meaning, not keyword density. Write naturally and factually.
• Including everything. The file should be concise — typically 50 to 200 lines. It is a summary, not a mirror of your entire website. Link to detailed pages for depth.
• Forgetting to update it. An outdated llms.txt with discontinued services or wrong pricing is worse than having no file at all. Set a quarterly review cadence at minimum.
• Using HTML or complex formatting. Stick to standard Markdown. HTML tags, images, and embedded media add noise that AI parsers may not handle reliably.
• Wrong file location. The file must be at /llms.txt on your root domain. Placing it at /about/llms.txt or /llms means AI systems following the convention will not find it.
• Missing links. Every service and content item should include a URL. AI agents use these links to dive deeper when they need more detail about a specific offering.
• Marketing language. Phrases like “best-in-class” or “industry-leading” are meaningless to AI agents. State facts: what you do, where you do it, what it costs.

Should you also create llms-full.txt?

The llms.txt specification proposes an optional llms-full.txt file for expanded content. While llms.txt is a business card, llms-full.txt is the full dossier — comprehensive Markdown versions of your most important pages.

Creating llms-full.txt is worthwhile if:

• You have complex service offerings that cannot be adequately summarized in a few lines.
• You publish technical documentation or detailed product specifications.
• You want to ensure AI systems have access to comprehensive, accurate descriptions of your products.
• Your industry requires detailed credentials, certifications, or regulatory information.

For most small-to-medium businesses, a well-crafted llms.txt is sufficient to start. Add llms-full.txt as your AI visibility strategy matures and you have data on which questions AI agents struggle to answer about your business.

Maintenance and long-term strategy

Your llms.txt file is a living document. Here is a practical maintenance cadence:

• Monthly: Verify all links are live and returning 200 status codes. Check that pricing and service descriptions remain accurate.
• Quarterly: Review the full file for outdated information. Add new services, content, or key facts. Remove discontinued offerings.
• After major changes: Any time you launch a new service, change pricing, move locations, or publish a significant piece of content, update llms.txt immediately.
• Annually: Audit how AI agents represent your business. Ask ChatGPT, Perplexity, and Gemini about your company and compare their responses to your llms.txt. Identify gaps and update accordingly.

Key takeaways

• llms.txt is a Markdown file at your domain root that provides AI systems with a structured summary of your business — proposed by Jeremy Howard of Answer.AI.
• It complements robots.txt and sitemap.xml as part of a complete AI visibility stack.
• Include your business identity, services, pricing, key facts, and links to important content.
• Keep it concise (50 to 200 lines), factual, and current. Avoid marketing language and keyword stuffing.
• Deploy at https://yourdomain.com/llms.txt and verify accessibility with a browser check and HTTP header inspection.
• In the emerging B2A2C model, llms.txt helps AI agents quickly evaluate and recommend your business with accurate information.
• Review and update the file at least quarterly to prevent stale data from harming your AI reputation.
• Start with llms.txt, and consider adding llms-full.txt once you have a mature AI visibility strategy.