What structured content is and why it matters
Unstructured content is a blob of text on a page. Structured content separates content from presentation — breaking it into typed, labeled components (title, summary, body, author, category, related items) that can be assembled, filtered, reused, and delivered across channels without manual reformatting. A product description that's structured content can appear on the website, in a mobile app, in an email, in search results, and in an AI assistant's response — each formatted appropriately for the channel, from a single source of truth.
In an AI-driven world, structured content is not optional. AI systems — search engines, assistants, agents — can only reliably extract and act on content that is explicitly labeled and semantically organized. Unstructured pages get summarized poorly; structured content gets parsed accurately.
Content modeling
Content Modeling Workshop Core Method
Use when: designing a new CMS, migrating to structured content, or planning content for a new product surface.
A content model defines the types of content in your system and their relationships. Content type definition: Identify each distinct type of content (Article, Product, Help Article, Case Study, Person, Event). For each type, define its fields: which are required, what data type each accepts (text, rich text, image, reference to another type, date, enum), and field-level constraints (character limits, allowed values). Relationships: Define how content types relate: a Case Study references a Product and a Person (the customer). An Article belongs to a Category and has Tags. These relationships enable navigation, filtering, and cross-referencing. COPE principle: Create Once, Publish Everywhere. If the same content appears in multiple places (a product name on the website, in the app, in emails), it should be authored once and referenced, not duplicated. Workshop format: Bring content creators, designers, engineers, and product managers together. Map existing content to types, identify fields, and model relationships on a whiteboard before touching a CMS.
Metadata and taxonomy
Metadata Schema Design Core Method
Use when: defining the labels, categories, and tags that make content findable, filterable, and machine-readable.
Metadata is data about content — the labels that make content discoverable. Descriptive metadata: Title, description, author, date, content type. Used for display and search. Administrative metadata: Owner, last reviewed date, review cycle, status (draft, published, archived). Used for governance. Structural metadata: Category, tags, related content, content model type. Used for navigation and filtering. Design principles: use controlled vocabularies (predefined lists, not free-text) for categories and tags to prevent inconsistency. Make metadata entry as automated as possible — derive dates, authors, and categories from the authoring context rather than requiring manual entry. Validate metadata completeness on publish — incomplete metadata means unfindable content.
Taxonomy Architecture Core Method
Use when: organizing content into categories that support navigation, filtering, and discovery.
Taxonomies are the classification systems that organize content. Three types: Flat: A simple list of categories with no hierarchy (tags). Good for cross-cutting attributes (topic, audience, format). Easy to manage but doesn't express relationships. Hierarchical: Tree structure with parent-child relationships (Products > Software > Design Tools). Good for browse navigation but forces content into a single path. Faceted: Multiple independent classification dimensions applied simultaneously (topic + audience + format + difficulty). Most flexible — enables users to filter by any combination. More complex to design and maintain. Match taxonomy type to use case: hierarchical for primary navigation, faceted for search/filter interfaces, flat for supplementary tagging.
AI-ready content architecture
Content Architecture for AI Systems Framework
Use when: ensuring your content can be accurately parsed, retrieved, and used by AI search engines, assistants, and agents.
AI systems consume content differently than humans. Optimize for both: Schema.org markup: Add structured data (JSON-LD) for products, articles, FAQs, events, people, and organizations. This tells AI systems exactly what your content represents. Clear heading hierarchy: AI systems use headings to understand content structure. H1 for the page topic, H2 for major sections, H3 for sub-sections — never skip levels. Self-contained sections: AI retrieval systems (RAG) often extract sections, not full pages. Each section should make sense independently, with its own heading and enough context to be understood without the surrounding content. Explicit facts over implications: AI systems parse explicit statements better than implied meanings. "Our product costs $49/month" is AI-readable; "affordable pricing" is not. Content API: For programmatic access, expose content through a well-documented API with consistent field names, filtering, and pagination. APIs serve both AI agents and partner integrations.
Real-world examples
Case study
NPR COPE: create once, publish everywhere
NPR pioneered the COPE model — creating content in a structured CMS that separates content from presentation. A news story authored once is automatically formatted for the website, mobile app, smart speaker, and API consumers. Each output channel gets the fields it needs (the website gets the full story with images; the smart speaker gets a headline and audio summary). This approach lets NPR serve dozens of channels without multiplying content production effort.
Why it works: Structured content types with clear field definitions make multi-channel publishing automatic rather than manual. The content model does the work.
Common pitfalls
Over-modeling
Creating 50 content types with 30 fields each before publishing anything is paralysis. Start with the 3–5 most common content types, build them well, and add complexity as real needs emerge. A content model is never finished — it evolves with the content it supports.