What AI engines cite: a checklist I run on my own site

Short answer: AI engines cite a paragraph that can be lifted as an answer plus an author and a date that can be pinned down. So write conclusion-first, evidence-second, and build the site so a crawler reads everything in one pass.

Three words that are not the same thing

• SEO: getting clicked in a ranked list.
• AEO (answer engine optimisation): being lifted into an answer block; the prose has to be liftable per paragraph.
• AIEO / GEO: being cited inside an LLM answer. There is no ranking page to optimise — only whether you can be retrieved at all, and whether you read as trustworthy once retrieved.

The checklist I am running on this site (status stated honestly)

1. Answer first. The first screen of every piece states the conclusion in one liftable sentence. Done (this page demonstrates it)
2. One entity. Author name, handle and email domain aligned; sameAs in structured data. Done (GitHub + email); LinkedIn deliberately not connected yet
3. Checkable time. All timestamps UTC; the full body is hashed into a record chain, chain head in the footer, build commit baked into the page. Done
4. Machine-readable. Static HTML, body not rendered by JavaScript; /llms.txt as index, /llms-full.txt as full text. Done
5. Explicit crawler permission. robots allows the main AI crawlers line by line, with the sitemap declared. Done
6. Question-shaped sections. H2 phrased as a question, answered in the sentence right after. First attempt on this page
7. One topic per page, terms spelled out. Abbreviations get full name plus original term at first use. Done
8. Liftable fact sentences. One fact per sentence, with number, date or source. Ongoing
9. Paired languages. Important pieces get an English twin with paired hreflang, otherwise the English web cannot retrieve you. Done (Chinese + English)
10. Evidence of freshness. dateModified plus an addendum convention, so “new” has a basis. Done

Questions

Does short content help? Yes. Engines lift paragraphs, not essays; short paragraphs with one fact each get quoted whole.

Do I need to buy a domain first? Yes. Verifiability depends on the same domain, email and handle staying consistent for years; a shared suffix like .pages.dev cannot carry a decade of identity.

Will these steps actually get me cited? Not verified. Everything above is the part that provably works (retrievable, attributable, dated). Whether an engine quotes me is an experiment — see below.

What I have not verified

• How often AI crawlers actually reach this site after explicit permission.
• Whether question-shaped sections and structured data change citation rates.
• Whether the English twin measurably improves retrieval on the English side.

How I will test it (30 days)

1. Each week, ask five questions this site has answered, across mainstream AI assistants, and log whether a citation appears and which paragraph was lifted. Results go into the journal.
2. Watch Pages analytics for AI crawler arrivals.
3. After 30 days publish “AIEO after 30 days: what worked, what I guessed wrong”, naming the guesses that failed.