Why it matters

Authority is structural, not just textual. When an AI system sees a pillar page on a core concept, surrounded by supporting pages on methodology, proof, and tools, all interlinked, it reads a coherent domain of expertise — and weighs the whole cluster more heavily than any single page. A lone blog post on the same topic, disconnected from anything else, signals a one-off rather than a knowledge base.

The hub-and-spoke model — pillar pages each supported by 3–5 related pages, connected by descriptive in-body links — is how you turn individual pages into an authority graph that compounds.

Your citation surface is not only your website [Industry study] Answer engines also cite Wikipedia, Reddit, review sites and press coverage about you. An industry study (Goodie’s AEO Periodic Table study) reports that third-party pages are frequently cited alongside, or instead of, a brand’s own site. Inventory those earned representations alongside your own pages, and record who in the organisation is accountable for each.

Size does not buy entry. Foundation Inc. — Reddit Cloud 100 (Nov 2026) scored 71 Forbes Cloud 100 B2B SaaS companies on Reddit community strength against citations in seven AI assistants and found near-zero correlation between brand valuation and how often AI cited them from Reddit; six communities produced 73% of the citations, distinguished by active discussion volume. For each priority unbranded category query, find the communities actually discussed and cited, and check whether you have a genuine presence there. One platform, one sector, 71 companies — the transferable finding is the negative one.

What good looks like

Concepts are organised into linked clusters with a clear taxonomy:

A master/pillar page for each core concept (e.g. "What is GEO?").
3–5 supporting pages per pillar — methodology, case studies, tools, schema.
A defined taxonomy: Concepts → Methods → Proof → Tools.
Every page links to 2–3 related entities with descriptive anchor text — never "click here".
All canonical content is reachable within three clicks from a topic hub; no footer-only links or orphan pages.

Step by step

Create a pillar page per core concept For each concept you want to own, build one authoritative master page — "What is GEO?", "The AI Knowledge Signal Framework". This is the hub everything else points to.
Add 3–5 supporting spoke pages Around each pillar, publish supporting pages that deepen it: methodology, case studies, tools, schema reference. Each spoke links back to the pillar and to relevant siblings.
Organise by a defined taxonomy Sort pages into a consistent taxonomy — Concepts → Methods → Proof → Tools. The same taxonomy should drive your navigation, your llms.txt grouping (Phase 1.3), and your entity map (Phase 4.3).
Link with descriptive in-body anchors From each page, link to 2–3 related entities using descriptive anchor text — "the AI Knowledge Signal Framework", not "click here". Descriptive anchors tell a model what the destination is about.
Keep canonical content within three clicks Ensure every canonical page is reachable within three clicks of a topic hub. Avoid footer-only links and disconnected topics — orphaned pages weaken the whole base (and the sitemap/llms.txt cross-reference, Phase 5.5).

Worked examples

1 Before / after: scattered posts vs. knowledge base

Before A single blog post on GEO; pages not linked to each other; footer-only links; "click here" anchors; disconnected topics; no central page. Authority can't accumulate.

After An "AI Knowledge Signal Framework" master page → a linked hierarchy (GEO → Claims → Schema → Case Studies), every page linking to 2–3 related entities with descriptive anchors, organised by Concepts → Methods → Proof → Tools.

2 A hub-and-spoke map

One pillar, its spokes, and the descriptive links that bind them.

PILLAR:  What is GEO?  (Concept)
  +- How GEO works            (Method)
  +- Citation study (yours)  (Proof)
  +- Free AI visibility audit (Tool)
  +- JSON-LD schema guide     (Method)

Each spoke links back to the pillar and to one sibling,
using descriptive anchors ("our 2026 citation study"),
all reachable within three clicks of the hub.

Common mistakes

Mistake	Fix
One isolated post per topic	Build a pillar with 3–5 supporting spokes.
"Click here" / "read more" anchors	Use descriptive anchor text naming the destination.
Footer-only or orphaned links	Link in-body; keep content within three clicks of a hub.
No taxonomy	Adopt Concepts → Methods → Proof → Tools consistently.
Pillars that don't link to spokes (or vice versa)	Interlink both directions to form a real cluster.
Assuming brand recognition earns third-party citations	Audit where your category is actually discussed and cited; brand size is not evidence of presence there.

Limits of this advice

Pillar size and click depth are heuristics, not thresholds any engine publishes.
A hub of thin pages is still thin. Each spoke must add something the pillar doesn’t (Phase 3.2).

Checklist

A pillar page exists for each core concept
Each pillar has 3–5 supporting spoke pages
Content is organised by Concepts → Methods → Proof → Tools
Every page links to 2–3 related entities with descriptive anchors
All canonical content is within three clicks of a hub — no orphans

How to Build a Knowledge Base (Hub-and-Spoke)