llms.txt Starter Template (2026): Copy-Paste Examples for Any Site
ChatGPT
Claude
Gemini
PerplexityThis page provides copy-paste llms.txt starter templates (basic, annotated, and per-site-type) aligned with the llmstxt.org spec. Each template uses the required H1 + blockquote + H2 sections of Markdown links so AI assistants like ChatGPT, Claude, Perplexity, and Cursor can ingest a site's most important pages without HTML parsing.
TL;DR. Save the template below as llms.txt, drop it at your domain root (https://yoursite.com/llms.txt), and replace the bracketed values with your site's name, summary, and top pages. The format is plain Markdown: H1 site name → blockquote one-line summary → H2 sections of - Title: description links.
This template is part of the Technical GEO reference hub and pairs with the llms.txt Reference (full spec) and How to Create llms.txt (step-by-step).
Spec Recap (Required vs Recommended)
The llms.txt format was proposed by Jeremy Howard at Answer.AI in September 2024 and is hosted at llmstxt.org. The spec is intentionally lean.
| Element | Status | Purpose |
|---|---|---|
| # Site name (H1) | Required | Must be the first non-empty line |
| > One-sentence summary (blockquote) | Strongly recommended | Often becomes the model's mental model of the site |
| Free Markdown text | Optional | Project notes, instructions, scope |
| ## Section + bullet links | Strongly recommended | Grouped links to detail pages, ideally .md versions |
| ## Optional section | Optional | Lower-priority links the AI can skip |
A companion file, llms-full.txt, may contain the full Markdown of every listed page in one document for deeper ingestion. Most sites only need llms.txt; documentation-heavy products often publish both.
Basic Template
Copy this, replace the bracketed values, and save as llms.txt:
txt
[Your Site Name]
[One-sentence description of what your site is and who it serves].
About
[2-3 sentences describing what your site offers, your audience, and what makes you authoritative.]
Key Pages
- Page Title: One-line description
- Page Title: One-line description
- Page Title: One-line description
Documentation
Topics Covered
- Topic 1
- Topic 2
- Topic 3
Contact
- Website: https://yoursite.com
- Email: contact@yoursite.com
Optional
- Less critical page: Description
Annotated Template
This version explains every line so you can adapt with confidence.
txt
Geodocs.dev
↑ H1 = your site or product name. Must be the first non-empty line.
Geodocs.dev is the canonical knowledge system for GEO and AEO.
↑ One-sentence blockquote. State what the site IS — not marketing copy.
About
Geodocs.dev provides reference documentation for Generative Engine
Optimization (GEO) and Answer Engine Optimization (AEO). It serves
SEO specialists, developers, and content strategists who need to
optimize content for AI search systems like ChatGPT, Perplexity,
Google AI Overviews, and Claude.
↑ 2-3 sentences. Audience + value + scope of authority.
Key Pages
- What Is GEO?: Canonical definition of Generative Engine Optimization
- GEO vs SEO: How AI search optimization differs from traditional SEO
- llms.txt Reference: Full specification reference
↑ Highest-value entry points. Quality > quantity. Format is strict:
"- Title: description"
Documentation
- How to Create llms.txt: Step-by-step tutorial
- Structured Data for AI: Schema implementation
↑ H2 sections group related pages. Use clear category names.
Topics Covered
- Generative Engine Optimization (GEO)
- Answer Engine Optimization (AEO)
- AI search technical standards
- Content strategy for AI visibility
↑ High-level topical map. Helps AI categorize the site.
Optional
- Glossary: Quick term definitions
↑ Optional sections signal lower-priority content the AI can skip
when context budget is tight.
Templates by Site Type
The same skeleton adapts to most site categories. Pick the closest match and edit.
SaaS Product
txt
[Product Name]
[Product] is a [category] that helps [audience] [primary outcome].
Product
- Features: Core capabilities and use cases
- Pricing: Plans and limits
- Integrations: Supported platforms
Documentation
- Getting Started: 5-minute setup guide
- API Reference: Endpoints and schemas
- SDK Docs: Language-specific clients
Resources
E-Commerce
txt
[Store Name]
[Store] sells [product category] for [audience], shipping to [region].
Categories
- Category 1: Description
- Category 2: Description
Shopping Help
- Shipping: Delivery options and timelines
- Returns: Return and refund policy
- Size Guide: Sizing tables
- FAQ: Frequent customer questions
Brand
- About: Company story and values
- Sustainability: Sourcing practices
Blog or Publisher
txt
[Publication Name]
[Publication] covers [topic area] for [audience], publishing [cadence].
Featured Content
- Cornerstone Article 1: Description
- Cornerstone Article 2: Description
Topics
- Topic 1
- Topic 2
- Topic 3
About
- Editorial Standards: How we report and verify
- Team: Writers and editors
- Contact: Pitch and tips line
Documentation Site
txt
[Project Name] Docs
Official documentation for [project], a [one-line description].
Get Started
- Installation: Setup in under 5 minutes
- Quickstart: Hello-world example
Guides
- Authentication: How to authenticate
- Deployment: Production deployment
Reference
- API: Full API reference
- CLI: Command-line reference
- Configuration: All config options
Optional
- Changelog: Version history
- Migration guides: Upgrade paths
Deployment Checklist
- Save as llms.txt — plain text, UTF-8, no extension changes.
- Upload to your site root — must be reachable at https://yoursite.com/llms.txt, not under a subdirectory.
- Verify in a browser — visit the URL; you should see raw Markdown, not an HTML page.
- Link .md versions when possible — many AI tools prefer page.md over page so they receive clean Markdown without HTML parsing.
- Reference from robots.txt — add a comment like # llms.txt: https://yoursite.com/llms.txt to aid discovery.
- Update on significant changes — when you add or restructure key pages.
- Optionally publish llms-full.txt — a single Markdown bundle of every listed page for deeper ingestion.
Validation
After deploying, run a quick three-step validation:
- Fetch test. curl -I https://yoursite.com/llms.txt should return 200 OK with Content-Type: text/plain (or text/markdown).
- Render test. Open the URL in a browser; the H1, blockquote, and link lists should be readable plaintext.
- Agent test. In an AI tool with web fetch (Claude, ChatGPT, Cursor), ask: "Read https://yoursite.com/llms.txt and summarize the site." A good template returns an accurate summary on the first try.
Common Mistakes
- No H1 on the first line. The spec requires the H1 to come first; comments or blank lines before it break parsers.
- Marketing language in the blockquote. AI systems often quote this line verbatim — keep it factual and specific.
- Listing every page. llms.txt is curated, not a sitemap. Include only entry points.
- Linking only to HTML. When a .md version exists, link to it for cleaner extraction.
- Forgetting to update. Stale templates point to dead URLs and erode trust over time.
FAQ
Q: Where exactly does llms.txt go?
At your domain root: https://yoursite.com/llms.txt. Subdomains use their own root (https://docs.yoursite.com/llms.txt). Subdirectory placement (/blog/llms.txt) is not part of the spec, and most AI tools will not look for it.
Q: Do I need both llms.txt and llms-full.txt?
Most sites only need llms.txt. Documentation-heavy products (SaaS docs, API references) often benefit from publishing llms-full.txt as well, which concatenates the full Markdown of every listed page so AI assistants can ingest the whole product in one fetch.
Q: Will llms.txt improve my Google rankings?
No. Google has stated publicly that its search systems do not read or act on llms.txt as of 2025-2026. Its primary value today is for AI coding assistants (Cursor, GitHub Copilot, Claude) and AI search engines that opt in to the format.
Q: Does llms.txt block crawlers like robots.txt?
No. llms.txt is a discovery and indexing aid — it tells AI systems what to read, not what to avoid. Use robots.txt for blocking and ai.txt for AI-specific access policies.
Q: How long should the template be?
Aim for under ~200 lines. The point is curation: a focused list of entry points. Sites that try to list every URL defeat the purpose, since AI tools then waste context window on low-value pages.
Q: Can I use the same template across subdomains?
Each subdomain needs its own llms.txt because crawlers fetch the file relative to the host. A marketing site, docs subdomain, and app subdomain typically each get their own tailored version.
Related Articles
ai.txt: AI Agent Access Policy Reference
ai.txt is an emerging root-level file that declares site-wide permissions and attribution rules for AI training, citation, and inference.
How to Create llms.txt: Step-by-Step Tutorial for AI Search
Step-by-step tutorial for creating, deploying, and validating an llms.txt file so AI systems and LLMs can discover your site's most important content.
llms.txt Reference: Specification, Format, and Examples
llms.txt is a proposed root-level Markdown file that gives LLMs a curated, machine-readable index of a site. Reference for spec, format, and adoption.