Skip to content

Pillar 3 — Answer-Engine Quality

Layer B — GEO Visibility. Owner: Content + SEO. Weight: 30 pts.

Highest-weighted pillar. No public scanner covers it — Cloudflare’s IsItAgentReady scores Pillars 1, 2, 4, 5 but not this one. Pillar 3 is where citation share is actually won.

#ItemStatusAuthority
10Quick-answer paragraph (40–80 words)framix conventionHouse rule
11Schema.org JSON-LD (Organization + WebSite + Article / Product / Service + BreadcrumbList; Person schema on every byline)requiredschema.org
12Author byline + bio + LinkedIn + credentials; primary-source citations on data-driven piecesrecommendedE-E-A-T convention
13Freshness + deprecation governance (dateModified; noindex + banner + canonical + Link: rel="successor-version" on deprecated posts; monthly scan flags > 12 mo)recommendedHouse rule informed by Cloudflare Redirects for AI Training behaviour

Full item descriptions: Universal Requirements § Pillar 3. Writing-side guidance in Content Playbook.

Triaged 2026-06-03. Decided items will land in the next revision; the open item stays under review.

  1. Add off-site authority as a Pillar 3 sub-component. Lily Ray, Kevin Indig, and SEL omnichannel-presence findings converge: third-party mentions are a primary driver of AI citations (_sources/expert-lily-ray.md). v1.1 scores on-site only; v1.2 will add an off-site authority item to Pillar 3 — proposed weight 5 pts, redistributed from existing P3 items (final split deferred to the v1.2 spec). The brand-mention surface will be defined in Content Playbook alongside the block palette.
  1. Schema is not a direct AI-citation lever — closed in favour of v1.1’s current position. Ahrefs’ March 2026 difference-in-differences study (_sources/ahrefs-schema-ai-citations.md) shows no causal lift on already-cited pages. Schema stays required for rich results, voice, and Knowledge Graph eligibility — explicitly not claimed as a citation-share lever. Volpini and Barnard’s “structure is the moat” framing absorbs into Pillar 2 + Pillar 3 on-site discipline.
  1. Is “training-data presence” a scorable layer distinct from the RAG layer? Britney Muller frames it as a separate, longer-horizon lever (_sources/expert-britney-muller.md). v1.1 measures the RAG layer only. Deferred until a measurement surface exists — currently no public attribution path from a model output to a training-data document.

Cross-source comparison: _sources/expert-comparison-index.md.

Gate B — Content checks all six content items (B.1–B.6) against this pillar before any property ships.