llms.txt Generator
Build a spec-compliant llms.txt for your site — the markdown manifest that tells ChatGPT, Claude, Perplexity and other LLMs which pages to read. Fill the form, copy or download. Everything runs in your browser; nothing is uploaded.
How it works
The llms.txt standard, proposed by Jeremy Howard at Answer.AI on September 3, 2024, defines a small markdown grammar that an AI tool can read quickly to understand a site. This generator assembles that file deterministically — it is string templating, not a calculation — in the exact order the specification at llmstxt.org mandates:
- H1 title (required). Your project name becomes the first line:
# Name. The file is invalid without it. - Summary blockquote (optional). A single line prefixed with
>giving a one-sentence description of the site. - Detail block (optional). Free-form markdown for extra context. Headings are not allowed here — the spec reserves
#and##for the title and section names, so the tool flags any heading you paste in. - Sections. Each becomes a
## Sectionheader followed by a markdown list. Every link is rendered as- [title](url), with an optional: notesuffix. Links missing a title or a valid http(s) URL are left out and reported as errors. - Optional section (special). If enabled, the
## Optionalblock is always written last — these are links an LLM may skip when it needs a shorter context. - Trailing newline. The file ends with a single newline, and blocks are separated by one blank line.
To prove the output really matches the grammar, the generated text is fed back through an independent parser based on the reference repository at AnswerDotAI/llms-txt. When the parser recovers the same H1, the same section count and the same number of links the generator wrote, the file is marked verified — a round-trip check, not a guess. The tool also surfaces soft warnings (for example, a summary over 200 characters) without changing your text, so you stay in control of the final file.
Worked examples
Example 1 — minimal valid file
- →Name: Ryzera Docs
- →Summary: Developer documentation for Ryzera Technologies APIs.
- →Section “Docs” → link “Quickstart” → https://ryzera.lk/docs/quickstart
# Ryzera Docs > Developer documentation for Ryzera Technologies APIs. ## Docs - [Quickstart](https://ryzera.lk/docs/quickstart)
H1 present, blockquote present, one ## section, one valid link item, single trailing newline. Spec-compliant.
Example 2 — detail, a note, and an Optional section
- →Name: induwara.lk
- →Summary: Free Sri Lanka-specific calculators and developer tools.
- →Detail: Built by Induwara Ashinsana. No signup, no ads.
- →Section “Tools” → two links (first has note “IRD 2025/26”)
- →Optional section → link “About”
# induwara.lk > Free Sri Lanka-specific calculators and developer tools. Built by Induwara Ashinsana. No signup, no ads. ## Tools - [Income Tax Calculator](https://induwara.lk/tools/sri-lanka-tax-calculator): IRD 2025/26 - [GPA Calculator](https://induwara.lk/tools/sri-lanka-gpa-calculator) ## Optional - [About](https://induwara.lk/about)
The note suffix applies only to the first link; the Optional section is written last regardless of where you add it.
Example 3 — edge case: heading in detail and an incomplete link
- →Name: My API
- →Detail contains a line: # Reference
- →Section “Docs” → link with title “Guide” but a blank URL
# My API # Reference
The tool flags two errors — a heading inside the detail block (# Reference) and a link with no URL — and the badge stays amber until both are fixed. Because its only link is incomplete, the Docs section is left out entirely rather than written as an empty section or broken markdown.
Frequently asked questions
Sources & references
- llmstxt.org — the official llms.txt specification
- AnswerDotAI/llms-txt — reference grammar and example files
- Anthropic docs — a real-world llms.txt used to sanity-check formatting
The output format on this page was last cross-checked against the llmstxt.org specification and the AnswerDotAI reference files on 2026-06-28. The format is reviewed whenever the specification changes.
Related tools
Comments & feedback
Spotted a bug or want an improvement? Tell us — our team reviews every comment, and good ideas get built. Comments are public and anonymous.
Found a bug, edge case, or want to suggest an improvement?
Email me at [email protected] — most fixes ship within 24 hours.