AEO citability

Answer engines (Perplexity, ChatGPT search, Google AI Overviews) cite pages they can parse cleanly: pages with structured data, a plain-text version, and a machine-readable map of the site. This skill adds those to ordinary static HTML. It is Python 3 standard library only — no build system, no framework, no network calls — and every write is idempotent, so it is safe to run repeatedly.

The worker is {baseDir}/scripts/aeo.py. Always run audit first, then apply the fixes the audit reports, then audit again to confirm.

When to use this

Use it when the user wants their website, blog, docs or landing pages to be cited inside AI-generated answers, asks about "AEO", "answer engine optimisation", "getting cited by Perplexity/ChatGPT", "structured data / JSON-LD", "llms.txt", or generally wants their static site to be more machine-readable.

It works on any directory of .html files (a static site, an exported blog, a docs folder). It does not need their CMS or framework.

How to use it

1. Audit first

python3 {baseDir}/scripts/aeo.py audit <dir-or-files>

Prints a 0–100 readiness score per page across seven checks (title, meta description, canonical URL, og:image, Article JSON-LD, FAQPage JSON-LD, clean .md sibling) and lists the biggest wins. Lead with this so the user sees the gap.

2. Add Article / BlogPosting structured data

python3 {baseDir}/scripts/aeo.py schema <dir-or-files> \
  --type BlogPosting --author "Jane Doe" --publisher "Acme" \
  --base https://acme.example --date 2026-06-30

Derives the JSON-LD from each page's own <title>, meta description, canonical, og:image and dates, and injects it before </head>. Skips pages that already have Article/BlogPosting/NewsArticle schema (override with --force). Use --type NewsArticle for news, Article for general pages. Always show --dry-run output first if the user is cautious.

3. Add an FAQ (the block answer engines quote most)

python3 {baseDir}/scripts/aeo.py faq <file.html> \
  --qa "What is X?::X is ..." "How much does it cost?::It costs ..."

Injects FAQPage JSON-LD from Question::Answer pairs. Keep answers self-contained (no "this"/"it" referring back to the page) — that is what gets extracted into an answer. One file at a time.

4. Write clean .md siblings

python3 {baseDir}/scripts/aeo.py md <dir-or-files>

Writes page.md next to each page.html: title, description, source link, then the readable body as Markdown (nav/header/footer/script stripped). Many engines prefer a clean text version when one exists at the same path. Skips existing siblings unless --force.

5. Generate llms.txt — the map engines fetch

python3 {baseDir}/scripts/aeo.py llms <dir> \
  --base https://acme.example --site "Acme" --summary "What Acme publishes."

Writes llms.txt listing every page with its title and description. This is the file Perplexity and others fetch to decide what is worth citing. Put it at the site root.

6. Generate sitemap.xml

python3 {baseDir}/scripts/aeo.py sitemap <dir> --base https://acme.example

Standard sitemap from the HTML files present. Skip if the site already has one from its framework.

Notes for the agent

• Reference the script as {baseDir}/scripts/aeo.py — never hardcode a path.
• These commands edit the user's HTML in place (schema, faq) or write new files (md, llms, sitemap). If the user has not asked you to modify files, show --dry-run first and get a go-ahead.
• Re-running is safe: everything is idempotent.
• After changes, suggest the user submit the new sitemap.xml / llms.txt and ping IndexNow or Search Console so engines re-crawl.
• Built by Workloft (https://workloft.ai). The approach is the one Workloft uses on its own Labs and Ships pages.