Citation-Ready Knowledge Base: Information Architecture Checklist

A citation-ready KB has a clear taxonomy, consistent page templates, descriptive anchors, breadcrumb context, freshness signals, and machine-readable cues. Each IA layer either makes it easier for an AI engine to cite the right page — or it doesn't.

TL;DR

If any IA layer is loose, AI engines lose confidence. Tighten taxonomy, enforce templates, name your anchors, mark up breadcrumbs, expose freshness, and standardize machine-readable cues. Use this checklist on the next KB audit.

How to use this checklist

The IA layers below are ordered by impact on citation. Run the checklist against:

A representative sample of 20-30 KB pages.
The KB's hub pages and category landings.
The KB index, sitemap, and llms.txt entries.

A pass is unanimous yes for that page; partial passes go on the remediation list.

Layer 1 — Taxonomy

[ ] Each page belongs to exactly one canonical category.
[ ] Categories are mutually exclusive.
[ ] Subcategories are consistent depth (≤3 levels recommended).
[ ] Each category has a hub page that summarizes coverage.
[ ] Tags are controlled vocabulary, not free text.
[ ] Synonyms and aliases are documented in a glossary.
[ ] No orphan categories (categories with one page only).

Layer 2 — Page templates

[ ] Every content type uses a documented template (guide, reference, comparison, framework, checklist, specification).
[ ] Templates open with an answer block within the first 60 words.
[ ] Templates include H2 anchors for major sections.
[ ] Templates include a FAQ section where appropriate.
[ ] Templates close with a related-references list.
[ ] Page bodies follow a consistent reading flow per content type.
[ ] Editorial style guide enforces tone, length, citation patterns.

Layer 3 — Anchors and headings

[ ] H1 matches the canonical question.
[ ] H2s use task-shaped phrasing ("How to…", "What is…", "When to…").
[ ] H3s never exceed three deep.
[ ] Anchor IDs are stable kebab-case slugs.
[ ] Internal links use anchor IDs where the section is the target.
[ ] Anchors do not change between revisions without a redirect.

Layer 4 — Breadcrumbs and navigation

[ ] Every page renders a breadcrumb mirroring its taxonomy path.
[ ] BreadcrumbList schema is present and valid.
[ ] Sidebar nav highlights the current section.
[ ] Hub pages link out to all child pages with descriptive anchor text.
[ ] Cross-section linking exists where genuinely useful.
[ ] No-broken-link policy enforced in CI.

Layer 5 — Freshness signals

[ ] Each page shows Last reviewed and Last updated dates.
[ ] dateModified schema reflects material changes only.
[ ] A periodic review cadence (e.g. 90 days) is enforced.
[ ] Outdated pages get a banner or are unpublished.
[ ] Changelog or version history visible on stable references.
[ ] Sitemap lastmod mirrors dateModified.

Layer 6 — Machine-readable cues

[ ] Article or TechArticle schema on every content page.
[ ] FAQPage schema on pages that genuinely have an FAQ block (see LLM-friendly FAQ schema).
[ ] sameAs schema links primary entities to Wikidata or canonical sources.
[ ] llms.txt lists Tier-1 KB pages.
[ ] sitemap.xml is segmented and current.
[ ] Open Graph tags accurate per page.
[ ] No JavaScript-only content for primary text — server-render or hydrate.

Layer 7 — Search and findability

[ ] On-site search returns relevant results for primary entities and aliases.
[ ] Empty-state queries get a useful response (suggest categories or popular pages).
[ ] Related pages section curated, not algorithmic-only.
[ ] Each page has at least 3 inbound internal links from related content.

Layer 8 — Authority signals

[ ] Author byline present per page.
[ ] Reviewed-by line included for YMYL or technical correctness.
[ ] About page documents editorial standards.
[ ] Citation hygiene rules enforced for sources.
[ ] Publication and review dates visible.

Validation

IA audit score ≥8/10 across all layers above.
90% of sampled pages pass each individual checklist line.
AI engine sample queries cite the correct hub or page (not a competitor).
Search Console errors for structured data trend toward zero.
30-day retest after IA changes shows uplift in citation share.

Common pitfalls

Free-text tags. Killing taxonomy.
Templates ignored. Editors override layout, breaking machine-readable cues.
Anchor drift. Anchor changes break inbound deep links.
Schema-body mismatch. Schema describes content not present in the rendered DOM.
No freshness pressure. Pages decay; reviewers do nothing.

FAQ

Q: How is this different from a traditional KB IA checklist?

The traditional checklist optimizes for human findability. This one adds machine-readable cues, freshness signals, and entity discipline that specifically influence AI citation.

Q: What's the minimum viable version of this?

Layers 1-3 (taxonomy, templates, anchors). Without them, the rest of the IA collapses regardless of schema.

Q: Do I need to ship llms.txt for KB content?

For citation-critical KBs, yes. The cost is low and the steering value for AI retrievers is meaningful.

Q: How often should we run the audit?

Full audit quarterly; sampled audit (10-20 pages) after every major release.

Q: Can I automate parts of this?

Yes. Schema validation, breadcrumb checks, anchor stability, and link checking are CI-friendly. Taxonomy and template adherence still need human review.