Answer Engine Optimisation

---
name: answer-engine-optimisation
description: Structure any page so answer engines can quote it and people still convert — page-type workflows, answer-landscape research, entity relationships, technical schema, and a scored AEO rubric. Use when asked to optimise for AEO or GEO, get cited by ChatGPT or Perplexity, show up in AI Overviews, make a homepage or landing page AI-quotable, adapt SEO for answer engines, balance AEO with conversion copy, or someone says 'AI answers never mention us'. For blog writing, see blog-content-writer. For conversion craft, see landing-page-copy. For scoring live posts, see seo-audit.
metadata:
  version: 1.1.0
---

# Answer engine optimisation

Google, ChatGPT, Perplexity and AI Overviews do not reward one clever article —
they reward **topical authority**, and at page level, content an engine can lift
*and attribute*. This skill is the build-side system for that: research what
answers are being cited, structure yours to be liftable, wire the entity and
technical signals, and score the result.

Two honesty lines up front: nobody can guarantee a citation — this builds
*quotable and attributable*; the engines choose. And AEO never overrides
conversion — **a page that gets cited but doesn't convert has optimised for the
wrong reader.**

## Before you start

| Input | Why |
| --- | --- |
| **The page type** — blog post, homepage, landing page, product/course page | The workflow branches on it; blog rules applied to a homepage produce robotic heroes |
| The target queries and question feed | AEO is question-shaped |
| The entity facts — canonical name, founder, location, category, credentials | Attribution needs relationships, not just repetition |
| The claims register with sources and dates | Engines quote sourced claims; **no invented proof, ever** |
| The site's cluster state | Page tactics ride on site authority |

## Step 1 — Set the goal balance by page type

Four goals compete on every page: **SEO** (rank) · **AEO** (extractable) ·
**Conversion** (act) · **Brand** (remember). The page type sets the priority:

| Page type | Priority order | What AEO looks like here |
| --- | --- | --- |
| Blog post / pillar | AEO = SEO > Brand | The full structure: 8–15 question H2s, FAQ block, liftable sections |
| Homepage | Brand = Conversion > AEO | One liftable "what is [entity]" definition + entity signals; the rest is positioning and proof — see landing-page-copy |
| Landing page | Conversion > AEO | Answer-first hero, one clean definition, sourced proof bar; persuasion leads |
| Product / course page | Conversion = AEO | Spec-shaped facts (price, duration, outcomes) in liftable blocks + Course/Product schema |

On non-blog pages, AEO earns its place only where it does not cost persuasion.

## Step 2 — Research the answer landscape

Structure without research optimises the wrong answers. Before writing:

1. **How people phrase it** — PAA/AlsoAsked mining and Search Console queries
   (via keyword-research), split by intent: branded / informational / commercial.
2. **Who is cited today** — ask the target questions in the actual engines
   (AI Overviews, Perplexity, ChatGPT) and record which entities get named and
   for what claim. That list is the competitive bar.
3. **What they have that you lack** — original data, primary sources, tables,
   named frameworks. Information gain is what flips citations; matching content
   earns paraphrase, exceeding it earns the name-check.
4. **Verify your own claims** — every number the page will carry gets its
   source and date *now*, before writing. Unverifiable claims are cut here, not
   caught later.

## Step 3 — Make the content liftable

- **Sections survive alone**: no "as mentioned above", no dangling pronouns.
- **Answer first, elaborate second** — the first two sentences under any
  heading carry a complete, correct answer.
- Blog posts: question H2s + FAQ block (the full blog-content-writer structure).
  Pages: each block self-contained; the hero answers "what is [entity]" in 2–3
  sentences before positioning language.
- **Definitions in one clean sentence**; comparison tables for choice questions
  — engines lift tables wholesale.
- **Claims quotable**: number + source + date, from the verified register.

## Step 4 — Build the entity by relationships, not repetition

Repetition is the weakest entity signal. The system:

- **One canonical name string** everywhere — never three variants.
- **Relationships stated and marked up**: founder ↔ organisation, location,
  category, credentials, courses/products — in the copy once, and in structured
  data permanently (Organization/Person schema, `sameAs` to real profiles).
- **Author and reviewer attribution** correct on every post.
- The corpus's **10–20 natural mentions applies to long-form posts only**
  (2,500+ words). On a homepage or landing page, mention density follows
  readability — a hero that says the brand ten times is spam to every reader,
  human or engine.

## Step 5 — Wire the technical layer

Recommendations by page type — the full table with schema types, `sameAs`,
canonicals, dates, crawlability and the llms.txt caveat lives in
[references/technical-aeo.md](references/technical-aeo.md):

| Page | Minimum schema | Also |
| --- | --- | --- |
| Blog post | Article + FAQPage (if FAQ) + Person (author) | dateModified real, canonical set |
| Homepage | Organization (or EducationalOrganization) + sameAs | One canonical brand definition |
| Course/product | Course/Product + Organization | Breadcrumbs; spec facts in markup |
| All | — | Server-rendered content; visible update dates; internal links from the cluster |

llms.txt: cheap to add, **impact limited and uncertain** — say so when
recommending it; never sell it as a lever.

## Step 6 — Score the pass (the rubric, not a checklist)

Rate each dimension 0–10 against its criterion; the report carries the scores
and the weakest two become the work order. A worked before/after example:
[references/before-after-example.md](references/before-after-example.md).

| Dimension | 9+ looks like |
| --- | --- |
| Intent coverage | Branded/info/commercial queries each answered somewhere on the page or cluster |
| Information gain | Something citable competitors don't have: original data, a named framework |
| Evidence strength | Every claim numbered, sourced, dated; zero invented proof |
| Entity clarity | Canonical name + marked-up relationships; a stranger could write the knowledge-panel entry |
| Citation readiness | Sections survive extraction; definitions one-sentence; tables where engines want them |
| Schema readiness | Page-type schema present and valid, not just recommended |
| Freshness | Real dateModified; stale facts absent |
| Conversion fit | The page still persuades — AEO cost the copy nothing it needed |

## Output

```
# AEO pass: <page> · type: <blog/homepage/landing/product> · v<page version>

Goal balance: <priority order applied>
Research: <n> query phrasings by intent · cited today: <entities+claims> ·
          gain opportunity: <what we have that they don't>
Claims: <n> verified (source+date) · cut as unverifiable: <n> · invented: 0

## Rubric
| Dimension | /10 | Evidence |
[8 rows] → Overall: <n>/10 · Work order: <weakest two, with fixes>

Entity: canonical "<name>" · relationships marked up: <which> · sameAs: <n>
Technical: <schema shipped vs recommended — gap named>
Honesty: quotable built; citation is the engine's call. llms.txt: <added, impact uncertain / skipped>
```

## When it breaks

| What you see | What it means | The fix |
| --- | --- | --- |
| Homepage reads robotic after the pass | Blog rules applied to a page type | Step 1's balance: brand and conversion lead on a homepage |
| Technically compliant, commercially flat | AEO structure without conversion craft | Pair with landing-page-copy; the conversion-fit dimension gates |
| Cited-competitor list never checked | Research skipped — structure optimised blind | Step 2 first; the citation bar is empirical |
| AI answers use your facts, name others | Weak entity relationships or unsourced claims | Step 4's relationship markup + sourced claims |
| Brand name 10+ times in a hero | The long-form rule misapplied | Scope: 10–20 is for posts; pages follow readability |
| Schema "recommended" forever, never shipped | The pass ended at advice | The rubric scores shipped schema, not intentions |
| Proof bar numbers unverifiable | Claims added at design time | Step 2.4: verify or cut before writing — never publish first |
| High rubric score, no citations after months | The buildable half is done; site authority may be thin | Check the cluster (site-architecture); then patience — the engine's half |

Never chase the rubric by weakening its criteria. A page scored 9 on an honest
rubric and cited nowhere still beats a fake 9 — the first is waiting on the
engines, the second is lying to you.

## Rules

- **Page type sets the workflow**, because blog structure on a homepage
  produces liftable copy that persuades nobody.
- **Research before structure**, because perfectly formatted answers to the
  wrong questions optimise nothing.
- **Relationships over repetition for entities**, because engines associate
  through marked-up connections; density is for long-form flow, not identity.
- **No invented proof, verified before writing**, because a fabricated number
  in a cited answer is your worst claim at its widest reach.
- **Conversion is a rubric dimension, not a casualty**, because the page's job
  is the visitor's action; the citation is a channel to it.
- **Never promise citations**, because quotable and attributable is the
  buildable half — chosen is the engine's half.

## Related skills

- **keyword-research** — the phrasing and intent mining step 2 runs on.
- **landing-page-copy** — the conversion craft non-blog pages lead with.
- **blog-content-writer** — writes the blog-type structure natively.
- **seo-audit** — scores live posts across the four frameworks post-publish.
- **site-architecture** — the cluster authority all of this rides on.