Answer Engineering

This skill structures content so answer engines can find the answer, lift it cleanly, and attribute it to you. Engines do not read a page top to bottom and reward length; they slice pages into passages and pull the one that best answers the query. If your answer is buried, hedged, or tangled with three other ideas, it does not get used — no matter how good the page is. Answer engineering is the craft of making each passage extractable. It works for any business, B2B or B2C, product or service.

This is about structure and clarity. The evidence that makes an answer credible (statistics, quotes, citations) is evidence-and-citations; the markup that helps machines parse it is structured-data; whether a crawler can read the page at all is ai-retrievability. Use them together.

When to use this skill

• Writing new content intended to be cited by AI.
• Rewriting content that ranks in search but is never quoted in AI answers.
• Structuring FAQs, definitions, comparisons, and how-tos for extraction.
• Auditing a page's structure for "AI-readability."

Before you write

Read aeo-foundation.md for the audience and the target prompts — you are writing answers to specific questions people ask AI, so start from those. If it is missing, point to the aeo-foundation skill. Confirm: the exact question this passage answers, and the one fact or recommendation the reader (and the engine) should walk away with.

Core principles

• Answer first (BLUF). Lead each section with the direct answer, then expand. Engines weight the opening of a passage heavily; the conclusion-first "inverted pyramid" beats the slow build-up.
• One question, one section. Each section answers a single question completely. Mixed ideas produce passages an engine cannot cleanly lift.
• Self-contained passages. Write each section so it makes full sense pulled out of the page with no surrounding context — because that is exactly how it will be used.
• Avoid the middle "dead zone." Models recall the start and end of long documents far better than the middle. Put key answers near the top of the page and near the top of each section, not buried mid-article.
• Specific and verifiable beats padded. A tight, concrete answer is extracted; filler and keyword repetition are ignored or penalized (see the shared citation ranking factors).
• Match how people ask. Phrase headings as the natural-language questions from the prompt set, not as keyword fragments.

The answer-block pattern

The reusable unit of AEO content:

1. Heading = the question, phrased the way a person would ask an assistant ("How much does X cost?" not "Pricing").
2. A 40–60 word direct answer immediately after the heading. Self-contained, no "as mentioned above," no setup. This is the part engines lift, so make it complete and accurate on its own.
3. Expansion — 150–400 words of context, evidence, examples, and nuance for the human reader and for follow-up depth.
4. A clean exit — a takeaway or transition; do not let the next idea bleed in.

See references/answer-patterns.md for worked before/after examples, the BLUF rationale, the dead-zone effect, and FAQ formatting.

Workflow

1. List the questions. From the foundation prompt set, gather the exact questions this page should answer. One page can answer several related questions, one per section.
2. Outline as questions. Turn each into a question-style heading; order them the way a reader's curiosity flows (often broad → specific).
3. Write the answer block for each: 40–60 word direct answer, then expansion.
4. Make each passage self-contained. Reread each section alone — does it stand without the rest of the page? Remove back-references.
5. Front-load the page. Put the most important answer (and a short summary) near the top; keep critical content out of the deep middle.
6. Format for chunking. Short paragraphs (2–4 sentences), lists for steps or sets, tables for comparisons, descriptive subheadings.
7. Hand off. Add evidence with evidence-and-citations, markup with structured-data, and verify crawlability with ai-retrievability.

Run the result against templates/answer-block-checklist.md before publishing.

Output format

• The restructured content, ready to publish, with question-style headings and answer blocks in place.
• For key sections, the 40–60 word answer block called out so the user can see and refine the exact extractable passage.
• A short note on what changed and why (e.g., "moved the pricing answer above the fold; split the buried comparison into its own section").
• Any (assumption) about the audience question or claim flagged for confirmation.

Common pitfalls

• Burying the answer under a long warm-up.
• Sections that answer three questions at once.
• Passages full of "as we saw above" that collapse when extracted.
• Important answers stranded in the middle of a long page.
• Question headings written as keyword stubs instead of real questions.
• Padding to hit a word count — it dilutes the passage and can hurt citation.

Tool

Score a page (Markdown, HTML, or URL) for extractability — question headings plus 40-60 word answer blocks:

python scripts/score_answer_blocks.py page.md
python scripts/score_answer_blocks.py --url https://example.com/page

Optional. The script only automates this step; it needs python3 (standard library only, no install). No Python? Just follow the steps above manually.

Related Skills

• aeo-foundation — the audience and the prompt set you are answering.
• evidence-and-citations — the statistics, quotes, and sources that make answers credible and cite-worthy.
• structured-data — schema (incl. FAQPage) that reinforces extraction.
• ai-retrievability — ensures crawlers can actually read the structured content.
• ai-visibility-audit — checks interpretability/extractability across the site.