Information architecture

Rules

Write the site map before building pages. Every page has one parent (for breadcrumbs) even if it belongs to many categories (through tags or taxonomies).
Every page is reachable within three clicks of the home page.
Hubs (category or family pages) exist before or with their children, so they are never half-empty.

Each thing has one canonical name: the most widely used, plain name. Other names are listed as aliases and mentioned naturally in the copy.
Slugs come from the canonical name: lowercase, hyphenated.
Names follow a consistent pattern within a collection (for example <qualifier> <thing>).

Categories, tags and relationships use controlled vocabularies defined in code (enums or reference collections), validated at build time.
A tag page is indexed only when it has enough items to be useful (default: 3 or more); otherwise noindex.
Relationships between items (related, prerequisite, next step, alternative) are stored as data and derived onto pages, never hand-maintained in two places.
Reciprocal relationships are checked automatically.

Direction	Meaning	Example
Up	To the parent hub	Breadcrumb + “See all …” link
Across	To siblings	Related items grid
Down	To details	Links to specific sub-pages, glossary terms

Link the first mention of another page in the body, not every mention.
Anchor text is the target’s name or a natural phrase containing it.
No orphans: every page has at least one internal link pointing to it; a build check reports orphans.
Footer links to the main hubs and trust pages (about, contact, privacy, editorial policy).

Visible breadcrumbs mirror the parent chain and match the BreadcrumbList structured data exactly.

Listing pages render all items as crawlable HTML; client-side filters only show and hide. Filter state lives in query parameters that are not indexed.