Jekyll-AEO

Answer Engine Optimization for Jekyll. Generates clean markdown copies of every page and produces llms.txt / llms-full.txt index files, following the llms.txt spec.

Add to your Jekyll site's Gemfile:

gem "jekyll-aeo"

Then run:

bundle install

That's it. With zero configuration, jekyll-aeo will:

1. Generate a .md companion for every HTML page (e.g., /about/index.html → /about.md)
2. Generate /llms.txt with an index of all pages
3. Generate /llms-full.txt with all page content concatenated
4. Inject <link rel="alternate" type="text/markdown"> tags into every HTML page

Run bundle exec jekyll build and check your output directory.

All settings are optional. Add to _config.yml. Configuration uses a strict schema — only keys defined in the plugin are accepted; typos or unknown keys are silently dropped.

jekyll_aeo:
  enabled: true                    # master switch; when false, all generation stops (default: true)
  exclude:                         # URL prefixes to skip
    - /privacy/
    - /error/
  dotmd:
    link_tag: "auto"               # "auto", "data", or omit to disable (default: "auto")
    include_last_modified: true    # add "Last updated" date to .md files (default: true)
    dotmd_metadata: false          # add YAML front matter block to .md files (default: false)
    md2dotmd:                      # source markdown → .md settings
      strip_block_tags: true       # strip comment/capture block content (default: true)
      protect_indented_code: false # protect 4-space indented code blocks (default: false)
    html2dotmd:                    # rendered HTML → .md settings (for plugin-generated pages)
      enabled: false               # convert rendered HTML to markdown (default: false)
      selector: null               # CSS selector (default: null — auto-detects main > article > body)
  llms_txt:
    enabled: true                  # generate llms.txt (default: true)
    description: ""                # override site description in llms.txt
    include_descriptions: true     # include page descriptions in llms.txt entries (default: true)
    front_matter_keys: []          # (reserved, not implemented)
    show_lastmod: false            # (reserved, not implemented)
    sections:                      # custom sections (auto-generated if omitted)
      - title: "Pages"
        collection: "pages"
      - title: "Products"
        collection: "products"
      - title: "Blog Posts"
        collection: "posts"
      - title: "Optional"
        collection: "profiles"
  llms_full_txt:
    enabled: true                  # generate llms-full.txt (default: true)
    description: ""                # override description in llms-full.txt header
    full_txt_mode: "all"           # "all" or "linked" (default: "all")
  url_map:
    enabled: false                 # generate URL map markdown table (default: false)
    output_filepath: "docs/Url-Map.md"  # output path relative to project root (default: "docs/Url-Map.md")
    show_created_at: true          # show generation timestamp in document header (default: true)
    columns:                       # columns to include in the table
      - layout
      - url
      - url_dotmd
      - dotmd_mode
      - excluded
      - path
      - page_id
      - lang
      - redirects
  robots_txt:
    enabled: false                 # generate robots.txt with crawler policy (default: false)
    allow:                         # search/retrieval bots to allow
      - Googlebot
      - Bingbot
      - OAI-SearchBot
      - ChatGPT-User
      - Claude-SearchBot
      - Claude-User
      - PerplexityBot
      - Applebot-Extended
    disallow:                      # training bots to block
      - GPTBot
      - ClaudeBot
      - Google-Extended
      - Meta-ExternalAgent
      - Amazonbot
    include_sitemap: true          # add Sitemap: directive (default: true)
    include_llms_txt: true         # add Llms-txt: directive (default: true)
    custom_rules: []               # additional bot-specific rules
  domain_profile:
    enabled: false                 # generate /.well-known/domain-profile.json (default: false)
    name: null                     # falls back to site.title or site.name
    description: null              # falls back to site.description
    website: null                  # falls back to site.url
    contact: null                  # REQUIRED — email or URL, no fallback
    logo: null                     # URL to logo image
    entity_type: null              # Organization, Person, Blog, NGO, Community, Project, etc.
    jsonld: null                   # custom JSON-LD hash to include

Control .md generation per page via dotmd_mode front matter:

---
title: Secret Page
dotmd_mode: disabled       # skip .md generation for this page
---

---
title: Blog
dotmd_mode: html2dotmd     # force HTML-to-markdown conversion
---

Available values: auto (default), md2dotmd, html2dotmd, disabled.

See lib/jekyll-aeo/generators/README.md for the full decision logic.

Pages with redirect_to in front matter are automatically skipped.

For every HTML page Jekyll renders, the plugin:

1. Re-reads the original source file from disk
2. Strips YAML front matter
3. Strips Liquid tags ({% %} and {{ }}) outside fenced code blocks; content inside {% raw %}…{% endraw %} is preserved (tags stripped, inner content kept)
4. Strips kramdown attribute annotations ({: .class}, {:width="300"})
5. Prepends the page title as an H1 header (if not already present)
6. Adds the page description as a blockquote (if present)
7. Writes the result as a .md file alongside the HTML output (for the root index, also writes index.html.md with the same content)

Pages generated by Jekyll plugins (e.g., jekyll-paginate, jekyll-archives) have no source file on disk and are normally skipped. With html2dotmd.enabled: true, the plugin converts their rendered HTML output to markdown instead:

jekyll_aeo:
  dotmd:
    html2dotmd:
      enabled: true
      selector: null               # optional CSS selector

The converter automatically extracts the main content area (<main>, then <article>, then <body>) and strips layout chrome (<script>, <style>, <nav>, <header>, <footer>). Set selector to a CSS selector (e.g., ".content", "#main") to target a specific element.

If your Jekyll site runs under a subpath (e.g., baseurl: /docs), all links in llms.txt will include the prefix automatically: /docs/about.md instead of /about.md. No additional configuration needed.

By default, jekyll-aeo injects a <link> tag into the <head> of every HTML page pointing to its markdown copy:

<link rel="alternate" type="text/markdown" href="/about.md">

This helps AI crawlers discover the machine-readable version of each page.

The link_tag setting controls this behavior:

With link_tag: "data", you can place the tag manually in your layout:

{{ page.md_link_tag }}

Or use the URL directly:

<link rel="alternate" type="text/markdown" href="{{ page.md_url }}">

Generated at the site root with:

• H1: site title
• Blockquote: site description
• Link to llms-full.txt for complete content
• H2 sections grouping pages by collection, with links to .md files

All individual .md file contents concatenated, separated by ---.

• full_txt_mode: "all" (default): includes every eligible page
• full_txt_mode: "linked": only includes pages that appear in llms.txt sections

Generate a markdown table of all HTML pages with metadata. Disabled by default.

jekyll_aeo:
  url_map:
    enabled: true

This writes a docs/Url-Map.md file (configurable via output_filepath) relative to your project root (the directory containing _config.yml when running Jekyll, or the source directory otherwise) — useful as a development reference that can be committed to version control.

The table is grouped by collection (Pages first, then alphabetically) with configurable columns:

Generate a /.well-known/domain-profile.json file following the AI Domain Data spec (v0.1). This provides AI assistants with authoritative identity metadata about your site. Disabled by default.

jekyll_aeo:
  domain_profile:
    enabled: true
    contact: "hello@example.com"   # required
    entity_type: "Organization"    # optional

The contact field is required — generation is skipped with a warning if not set. The name, description, and website fields fall back to site.title/site.name, site.description, and site.url respectively.

Valid entity_type values: Organization, Person, Blog, NGO, Community, Project, CreativeWork, SoftwareApplication, Thing.

You can include a custom JSON-LD object via the jsonld key:

jekyll_aeo:
  domain_profile:
    enabled: true
    contact: "hello@example.com"
    jsonld:
      "@type": "Organization"
      name: "Example Corp"

Generate a robots.txt that separates search bots (allowed) from training bots (blocked). Disabled by default.

jekyll_aeo:
  robots_txt:
    enabled: true

Default behavior: allows search bots (Googlebot, Bingbot, OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Applebot-Extended) and blocks training bots (GPTBot, ClaudeBot, Google-Extended, Meta-ExternalAgent, Amazonbot). Includes Sitemap: and Llms-txt: directives automatically.

If you already have a robots.txt file in your source directory, the generator skips and uses yours instead. Integrates with jekyll-sitemap — no conflicts.

Add custom rules for specific bots:

jekyll_aeo:
  robots_txt:
    enabled: true
    custom_rules:
      - user_agent: "SpecialBot"
        allow: "/public/"
        disallow: "/private/"

Add the {% aeo_json_ld %} Liquid tag to your layout to output structured data as <script type="application/ld+json"> blocks:

<head>
  ...
  {% aeo_json_ld %}
</head>

The tag automatically renders JSON-LD for 6 schema types based on your page's front matter and site config:

Add a faq: array with q: and a: pairs to your front matter:

---
title: FAQ
faq:
  - q: "What is Jekyll-AEO?"
    a: "A Ruby gem for Answer Engine Optimization."
  - q: "Does it work with jekyll-seo-tag?"
    a: "Yes, they cover different layers and don't conflict."
---

Add a howto: object with steps: to your front matter:

---
title: How to Install
howto:
  name: "Install Jekyll-AEO"
  description: "Steps to add Jekyll-AEO to your site"
  totalTime: "PT5M"
  steps:
    - name: "Add to Gemfile"
      text: "Add gem 'jekyll-aeo' to your Gemfile"
    - name: "Install"
      text: "Run bundle install"
    - name: "Build"
      text: "Run bundle exec jekyll build"
---

Add speakable: true to mark a page's title and first paragraph as voice-assistant-friendly:

---
title: About Us
speakable: true
---

The Article schema automatically skips when jekyll-seo-tag is installed, since seo-tag already outputs BlogPosting (a subtype of Article). All other schema types (FAQPage, HowTo, BreadcrumbList, Organization, Speakable) are always safe — they output different types that don't conflict. Multiple <script type="application/ld+json"> blocks per page are valid per the JSON-LD spec.

Controls how Liquid comment and capture blocks are handled.

With strip_block_tags: true (default):

Source:

{% comment %}
Internal note for editors.
{% endcomment %}

Welcome to the page.

Output .md:

Welcome to the page.

Comment and capture blocks are fully removed (tags + content), since they contain developer metadata, not page content.

With strip_block_tags: false:

The same source would produce:

Internal note for editors.

Welcome to the page.

Note: {% if %}, {% for %}, {% unless %}, and {% case %} blocks always preserve their content regardless of this setting, since the content between them is real page content.

By default, only fenced code blocks (``` and ~~~) are protected from Liquid/kramdown stripping. Enable protect_indented_code: true to also protect indented code blocks (4+ spaces after a blank line).

Recommendation: use fenced code blocks for code examples that contain Liquid syntax.

By default, llms.txt auto-generates sections by grouping pages by their Jekyll collection. To customize:

jekyll_aeo:
  llms_txt:
    sections:
      - title: "Documentation"
        collection: "pages"
      - title: "Products"
        collection: "products"
      - title: "Optional"          # LLMs can skip this section per spec
        collection: "profiles"

Use collection: null to match standalone pages (those not in any collection).

After building your site, verify AEO output with:

bundle exec jekyll aeo:validate

This checks:

• llms.txt exists and starts with an H1 heading
• llms-full.txt (if present) is non-empty
• All .md files referenced in llms.txt exist in the destination directory
• domain-profile.json (if present): valid JSON, required fields (spec, name, description, website, contact), and valid entity_type (invalid values emit a warning, not an error)

Respects baseurl when resolving file paths.

The following are automatically skipped (in order):

• Plugin disabled (enabled: false)
• Non-HTML outputs (CSS, JS, etc.)
• Pages with dotmd_mode: disabled in front matter
• Redirect pages (redirect_to in front matter)
• Documents in the assets collection
• llms.txt and llms-full.txt files
• Paths matching exclude prefixes
• Pages with no source file on disk (unless html2dotmd is enabled)

bundle install

Run unit tests:

rake test

Run all tests (unit + integration — builds the example site):

rake

Build the example site standalone:

rake site:build

Serve the example site locally:

rake site:serve

The integration tests build a full Jekyll site from demo/example.com/ and verify all generated outputs (llms.txt, robots.txt, domain-profile.json, markdown copies, JSON-LD schemas, link tags).

MIT. Copyright (c) 2026 ZAAI.