a builder's codex
codex · playbooks · AEO relevance engineering — passages, schema, mirrors, scorecard

AEO relevance engineering — passages, schema, mirrors, scorecard

Treat answer-engine optimization as an information-retrieval engineering discipline, not content production. The unit of citation is the passage, not the page. Each pass emits four artifacts side-by-side so the doc ships AI-ready, not just human-ready.

Source synthesis: Mike King (iPullRank, "relevance engineering" frame), Aleyda Solis (per-vertical/per-country shape; three-layer measurement — Measure AI search on three layers: Presence, Readiness, Business Impact), Lily Ray (YouTube + LinkedIn = highest-cited AI Overview sources; A single seeded fake claim can self-confirm in AI Overviews), Amanda Natividad (precision over breadth), Kevin Indig (citation rate vs. mention rate gap — Citation rate and mention rate are different metrics; comparative content closes the gap).

When to use

Per-vertical AEO/GEO pass on a single source artifact (vertical playbook, battle card, pricing page, integration page, help doc). Run before publishing the page; rerun whenever schema.org vocabulary shifts or competitor citation profiles change.

Inputs

Process

1. Read substrate. Positioning, ICP-anchor card, brand-voice, product-knowledge for any product entities in the target.

2. Extract canonical entities. Product names, integrations, competitors, ICP role titles, pricing tiers, capability claims. Output an entities.json with type tags compatible with schema.org Product / Service / Organization.

3. Produce 8–12 stand-alone passages, 40–90 words each. Each answers ONE buyer question. Each is self-contained — LLMs cite passages, not pages (Mike King's frame). Single buyer question per passage.

4. Generate FAQPage + HowTo schema blocks for the passages where the question/answer shape fits. Validate against the current schema.org spec.

5. Map each passage to a YouTube + LinkedIn long-form mirror. Lily Ray: those two surfaces are the highest-cited AI Overview sources. Output a mirror plan with title, opening hook, source passage.

6. Score on Aleyda's three layers (Measure AI search on three layers: Presence, Readiness, Business Impact):

- Presence: 10–15 buyer-prompt coverage list, recommendation-rate target, citation-rate target.

- Readiness: 10-trait checklist (accessible, transactable, navigable, factual, …) with current state per trait.

- Business Impact: keep three confidence layers separate — observed referrals / modeled lift / attributed deals. Never combine.

Outputs

Quality gates

Human checkpoints

Refusal patterns

Common failure modes

Open the interactive view →