Content Strategy Framework

The Tiered Content Framework

A content governance model five years in the making—now being pressure tested in the wild.
The relationship between atomic design principles and structured content has been a live conversation in the industry for years. Headless CMS practitioners, composable content advocates, and structured content strategists have all circled the same problem: if design systems can govern UI components with structural precision, why does content governance still operate through editorial guidelines and ad hoc conventions?

The Tiered Content Framework is my answer to that question—developed over five years of practice across enterprise, brand, and digital product engagements, and released publicly in April 2026.

What the framework contributes is not the observation that content needs structure. That observation is well-established. What it contributes is the operating model: a formal content architecture that gives every discipline—content strategy, information architecture, design systems, CMS architecture, engineering—a shared structural vocabulary, and makes content decisions traceable, scalable, and governed rather than ad hoc. The tier structure is the scaffold. The operating model is the point.

It extends Brad Frost's Atomic Design methodology into the content layer, governing semantic content objects—their meaning, structure, relationships, and governance rules—from the smallest structured field up to the full experience ecosystem. The fourth tier, Structures, draws directly from the work of former colleague Andrew Kaufman, whose model of content structures provided the foundational thinking this tier has since evolved from.

Since initial publication, practitioner feedback has driven three cross-cutting governance dimensions—The Intelligence Layer, Taxonomy, and The Machine-Legibility Layer—none of which were present in the first version. The framework is being pressure tested in the wild, and it is evolving accordingly.

The Tiered Content Framework is my attempt to fill the gap.

Most content programs fail the same way. New content gets created to fill gaps that already exist in the current inventory—just undiscovered. Quality drifts. Tone diverges. Strategy, audit, briefing, and creation run in disconnected tools with no shared data layer. Content gets created as editorial output rather than a governed, structured infrastructure. The Tiered Content Framework addresses this at the foundation—governing meaning from the smallest structured field up to the full experience ecosystem.

If you work in content strategy, information architecture, design systems, or enterprise digital product, I'd genuinely value your reaction. What holds up? What breaks? Where does the model not account for how your organization actually works?

The Framework

The Tiered Content Framework is an original content strategy operating model, developed as an extension of Brad Frost's Atomic Design methodology, specifically applied to content strategy, information architecture, and enterprise content governance.

Where Atomic Design governs UI components—atoms, molecules, organisms, templates, pages—the Tiered Content Framework governs semantic content objects: the meaning, structure, relationships, and governance rules behind every piece of digital content.

Disclaimer: The Tiered Content Framework extends the structural logic of Atomic Design as intellectual lineage. Brad Frost has not reviewed or endorsed it.

The Seven Tiers

Why It Matters

Most design systems address content through voice and tone guides or microcopy standards. What they don't address is how content should be structured, modeled, governed, or reused across an enterprise digital ecosystem. That gap is where the Tiered Content Framework operates.

The framework provides:

• A shared structural vocabulary across content strategy, information architecture, design systems, CMS architecture, and engineering

• Field-level modeling and reusable semantic objects—not just editorial guidance

• Intent-driven page architecture that connects user journeys to content hierarchy

• Governance from the field level up to the experience-ecosystem level

• A foundation for personalization, automation, omnichannel delivery, and AI-assisted content creation

• A governance scope that scales from a single content field up to an entire enterprise digital presence

What This Is

What This Looks Like in Practice
The tier names are precise by design—they need to be, because the framework is used by engineers building CMS architecture and by content strategists writing briefs. But the underlying idea is simpler than the vocabulary suggests.

Here's the short version:
Every piece of digital content is built from smaller pieces. The Tiered Content Framework names those pieces, defines what each one is responsible for, and gives every team involved—strategy, design, engineering, content—a shared way to talk about them.

Think of it like this:

1. A Quark is not content; it's the condition under which content is valid. A character limit. An approved terminology list. A tone parameter. A metadata schema rule. Quarks are invisible to users and authors in the final experience, but they're the reason a Particle governing a product name is correct every time it appears, across every system, regardless of who authored it or which AI generated it. Change a Quark, and you change everything downstream. That's why they carry the highest governance scrutiny in the entire framework.

2. A Particle is a single fact with a job. A product price. A button label. A street address. It's the smallest thing that can be governed—and governing it well means everything built from it inherits that quality.

3. A Cluster is a few facts assembled into something meaningful. An author card. A product teaser. An address block with a map link. It exists because those Particles belong together, and the combination means something a single field doesn't.

4. A Zone is a region of a page, or a section of an app screen, or a segment of a voice response, organized around what the user is trying to do at that moment. A trust-building section. A navigation area. A conversion prompt. The Zone governs what content belongs there and why.

5. A Structure is the full thing delivered to a user: a web page, an app screen, or an AI assistant answer. It's the complete, governed assembly of Zones—the content experience as the user encounters it.

6. An Ecosystem is a connected set of Structures unified around a domain, journey, or brand area. A hospital's oncology service line. A software product's help center. A brand's campaign architecture. The parts that belong together, governed together.

7. A Biome is everything—the complete digital content presence of an organization, across every channel and surface it maintains.

The governance rule that runs through all tiers is the same one a good editor applies instinctively: make the decision at the lowest level possible. Get the field right, and the component built from it is easier to govern. Get the component right, and the page section built from it requires less review. Let a bad Particle propagate upward, and you're correcting it in all places instead of one.

The framework makes that instinct operational—traceable, repeatable, and scalable across organizations where no single editor can review everything.

The Intelligence Layer

The tiers describe how content is structured, governed, and reused across a digital presence. But they were designed for a world where content is authored, published, and delivered statically. Agentic systems, conversational interfaces, and AI-driven personalization introduce a different challenge: content generated, assembled, and delivered dynamically, in real time, at scale.

The Intelligence Layer is not another tier. It is one of three cross-cutting governance dimensions—alongside Taxonomy and The Machine-Legibility Layer—each addressing a distinct aspect of how the tier structure operates in dynamic, AI-mediated environments. Where Taxonomy governs how content is classified and routed internally, and Machine-Legibility governs how content declares itself to external systems, the Intelligence Layer governs how each tier behaves when content is no longer static output but active, responsive, and machine-generated.

At the Particle level, it governs the structured fields and constraints fed into LLM prompts—ensuring that even dynamically generated content inherits brand-approved terminology, tone parameters, and field-level accuracy requirements.

At the Cluster and Zone levels, it governs how semantic objects and Zone-level compositions are assembled dynamically—ensuring that personalized experiences maintain structural coherence and intent alignment regardless of how they're generated.

At the Structure and Ecosystem levels, it governs the rules, guardrails, and audit mechanisms that determine what agentic content systems are permitted to produce—and what they aren't.

The practical implication: every governance decision in the Tiered Content Framework is also a prompt engineering decision. The more precisely an organization governs its content at each tier, the more reliably its AI systems will produce content that is accurate, on-brand, and structurally sound—without requiring human review of every output.

This is the governance foundation that intelligent experience systems require but rarely have.

Governance follows a single rule: govern meaning at the lowest tier possible; escalate only when structural impact demands it. A change to a Particle affects everything downstream. A change to a Biome requires executive-level governance across the entire content presence.

Taxonomy: The Attribute Layer

Taxonomy is the attribute layer that makes the tier structure machine-actionable. Classification is governed at the Quark level and declared at the Particle level—Quarks define which attribute values are valid; Particles carry them. From there, taxonomy cascades upward through the tier structure, validating zone-affinity rules, governing assembly decisions, and making personalization, routing, and dynamic content possible at every tier. Taxonomy classifies. Quarks constrain.

Taxonomy isn't another tier. It's a governance dimension that runs across all tiers, just as the Intelligence Layer does. The diagnostic failure mode it addresses is called static boxes. A well-governed tier structure, correct content at every Particle, coherent semantics at every Cluster, is still inert without taxonomy. There's no routing mechanism. No assembly logic. No way for an AI or personalization system to act on the structure. Without taxonomy, the tiers are a governance vocabulary. With it, they become a governed content system that can move, respond, and assemble dynamically.

At the Quark level, the valid taxonomy values are defined. Which Content_Types are approved for use, which Audience segments are in scope, which Intent values the system recognizes — these are Quark-level decisions. A Particle can't carry a Content_Type that isn't in the governed list; a Zone can't declare an affinity that hasn't been specified. Change a taxonomy value at the Quark level and every Particle, Cluster, Zone, Structure, Ecosystem, and Biome that depends on that classification is affected — the same propagation logic that governs every other Quark-level change.

At the Particle level, classification is declared. Each Particle carries its taxonomy attributes—Content_Type, Audience, Intent, and any other attributes governed by the Quark layer. A Particle without taxonomy attributes can be governed for meaning but cannot be routed, assembled, or personalized against. It's a correct, well-formed fact with no address in the content system.

At the Cluster level, taxonomy cascades upward, and zone-affinity rules begin to apply. A Cluster's combined classification, derived from its constituent Particles, determines which Zones it belongs to. A product teaser Cluster carrying Content_Type: product and Intent: convert belongs in a Conversion Zone, not a Trust Zone. Assembling Particles into a Cluster isn't just a semantic decision; it's a routing decision governed by taxonomy.

At the Zone level, taxonomy governs assembly. Zone-affinity rules specify which Clusters belong in which Zones based on taxonomy matches. A Zone's own classification, its topical scope, intent, and target audience, define what content it will accept. This is where personalization operates at the composition level: swap the Clusters in a Zone based on the Audience taxonomy, and the Zone serves a different segment without the Structure around it changing.

At the Structure level, taxonomy governs how Structures are assembled for specific audiences and intents. A product detail page, an editorial article, and a service landing page each have different taxonomy requirements that determine which Zones are assembled within them and in what arrangement. Structure-level taxonomy is also what makes automated content assembly tractable: an AI or CMS system can assemble a valid Structure because the taxonomy rules specify what belongs where.

At the Ecosystem level, taxonomy consistency determines whether the Ecosystem functions as a coherent content system. Personalization, related content, faceted navigation, and recommendations all depend on the consistent application of taxonomy across every Structure in the Ecosystem. An Ecosystem where taxonomy is applied inconsistently, some Particles classified, others not; the same kind of content carrying different Content_Types in different areas, is an Ecosystem where dynamic assembly cannot work reliably.

At the Biome level, cross-Ecosystem taxonomy consistency determines whether the full digital presence can be traversed as a coherent content graph. If taxonomy values diverge across Ecosystems, different Content_Types across brands or product areas, misaligned Audience segments, conflicting Intent values, the Biome-level content graph fragments at the boundaries. Personalization breaks where the taxonomies don't align. AI systems encounter inconsistent classification signals and resolve them badly.

The dependency the other two cross-cutting dimensions rely on is this: the Intelligence Layer can only govern how content behaves dynamically if the tier structure has something to act on, and taxonomy is that something. Machine-Legibility can only declare content's relationships and classification to external systems if those relationships and classifications have been established internally first. Taxonomy is the internal address structure that makes both possible.

The Machine-Legibility Layer

The Intelligence Layer governs how each tier behaves when content is dynamically generated, assembled, or delivered by AI and agentic systems. Taxonomy governs the attribute classifications that make the tiers machine-actionable internally. Neither addresses a third question that has become structurally consequential: how content declares itself to the systems that encounter it from outside.

Machine-Legibility Governance is the third cross-cutting dimension and layer of the Tiered Content Framework. It runs across all tiers—describing how each tier declares its identity, relationships, authority, and epistemic status to search engines, knowledge graphs, AI retrieval systems, and large language models.

Every content object exists in two registers simultaneously. It has a human-readable presentation—the text a person reads, the layout they navigate, the hierarchy they interpret visually. And it has a machine-readable declaration: the structured data, schema markup, metadata, and entity relationships that tell external systems what this content is, who it's for, and what it means.

The machine-readable register carries one further declaration that has become structurally consequential as retrieval systems harden inferred content into operational fact: the content's epistemic status. Every content object can declare itself as confirmed, inferred, unverified, or time-sensitive, and can declare the verification record (or its absence) attached to that status. This is distinct from authority declarations (who published it) and provenance declarations (how it was produced). It governs which external systems are permitted to treat as established fact and which they must surface as inference. Without this declaration, retrieval systems and LLMs cannot distinguish between a Particle that has been verified against current conditions and a Particle that was inferred at authoring time and never re-checked. The Machine-Legibility Layer is where this declaration belongs because the consequence is external: it shapes what downstream systems do with the content, not how the content is classified or generated internally.

These are not separate concerns maintained by separate teams. They are two expressions of the same governance decision. A content object that is governed in the human register but ungoverned in the machine register is only half-governed, and as AI-mediated surfaces become primary discovery channels, the ungoverned half increasingly determines whether the content is found, understood, and correctly represented at all.

At the Quark level, machine-legible governance is where the rules originate. Which schema properties are approved for use across the content system, which entity types are in scope, which structured data declarations are required versus optional, and these are Quark-level decisions. A Particle carries its machine-readable declaration because a Quark governs that it must, and constrains what it can say. Machine-legibility at every tier above is downstream of the governance decisions made here.

At the Particle level, machine-readable governance means each structured field carries not only its human-readable value but also its machine-readable declaration: the entity type it represents, the schema property it populates, and the relationships it holds with adjacent fields. A price field that renders correctly on screen but carries no schema markup is a governed Particle in the human register and an ungoverned Particle in the machine register. The content framework and the code content structure are the same governance decision expressed in two registers.

At the Cluster level, machine-legible governance means that semantic objects declare their internal structure and relationships among entities to retrieval systems. An author card that displays a name and bio to a human reader but carries no structured Person markup, no entity relationship to the content it authored, and no authority signals connecting it to external knowledge graphs is a Cluster that exists only in the human register.

At the Zone level, machine-legibility governance means page-area compositions declare their topical scope and intent to external systems. A Trust Zone that presents testimonials and credentials to a human visitor but provides no structured Review or Organization markup is a Zone whose persuasive architecture is invisible to every system encountering it through retrieval.

At the Structure level, machine-legibility governance means that page templates encode their full relational context: what the page is about, how it relates to other pages in its topic cluster, its canonical status, when it was last substantively updated, and the content type it represents. These are not metadata afterthoughts appended during a technical SEO pass. They are governance decisions that belong to the Structure's definition. A Structure template that does not specify its machine-readable declarations has deferred a governance decision to a team that may not have the content context to make it.

At the Ecosystem level, machine-legibility governance means the site's entity architecture, topical authority structure, and internal linking logic are coherent and machine-traversable. Orphan pages, inconsistent taxonomy application, and unrelated CMS field proliferation are not technical debt in this framing; they are governance failures at the Ecosystem level, because they prevent external systems from understanding how the parts of the content presence relate to one another and to the broader knowledge domain.

At the Biome level, machine-legibility governance means the organization's full digital presence, across domains, brands, and platforms, maintains a consistent entity identity and authority architecture that external systems can resolve. Conflicting entity declarations across properties, inconsistent Organization markup, and fragmented knowledge-graph signals degrade the Biome's legibility to every AI system attempting to understand what the organization is and what it has authority over.

Epistemic Status — Governed Schema
The four status values form the controlled vocabulary governing every epistemic declaration in a TCF-governed content system:

These four values are not a quality ranking. Inferred and unverified content is legitimate; the requirement is not that every input be confirmed before use, but that every input's status be declared before it travels. A system can operate on an unverified input. It cannot treat an unverified input as if it were confirmed.

The Schema
Five fields implement the declaration at the Particle level:

A note on confidence levels: speculative is intentionally not a governed value. A hypothesis or speculative claim is classified through Taxonomy, it carries a Content_Type designation, and holds inferred status with low confidence. Encoding content type into a verification-state field conflates two governance concerns that belong in separate layers.

How Status Propagates
Status inherits upward through the tier structure by the weakest-status rule: a Cluster's epistemic status is the weakest status of its constituent Particles; a Zone's status is the weakest of its Clusters; a Structure's status is the weakest of its Zones. Weakest follows the ordering unverified → inferred → time-sensitive → confirmed.

One case carries specific weight: a time-sensitive Particle whose recency window has expired propagates as unverified rather than time-sensitive. An expired verification is, by definition, no longer current—the Particle is now sourced but unchecked, which is what unverified means. Status can only travel back up once a new verification action occurs and is recorded.

AI-Generated Content
AI-generated Particles carry a second field alongside epistemic status: tcf:aiProvenance. It records the generation method, the model used, the inputs the model received, and whether a human reviewed the output before the Particle was governed. These are distinct fields because generation method and verification state are separate governance concerns. An AI-generated Particle that has been verified against current conditions carries Confirmed status. One that has not carries Unverified status. tcf:aiProvenance records how the content was made. tcf:epistemicStatus records whether it has been checked. Neither determines the other.

Alignment with the Agentic Accountability Playbook
The four status values are shared with the Agentic Accountability Playbook, which uses the same vocabulary to govern how agentic systems handle inputs as they travel through decision chains. The two frameworks govern different sides of the same exchange. The TCF governs the declaration side: what a content object asserts about its own verification state when it presents itself to an external system. The Playbook governs the receiving side: what a downstream node in an agentic decision chain is permitted to treat as confirmed. A content object can declare itself Confirmed under the TCF and still arrive in an agentic system as Unverified under the Playbook, because the agentic system has not re-verified it within its own recency window for its specific decision context. These are distinct governance claims, operating in the same space, by design.

The AI Search Fragment Problem
AI-mediated search surfaces—no-click answers, AI Overviews, retrieval-augmented generation, conversational search—introduce a specific failure mode this dimension must name.

When an AI search system extracts a content fragment and presents it as a standalone answer, it is performing a Particle-level extraction from a larger Structure. The fragment inherits none of the governance context that gave the original content its meaning: no source authority signal, no relationship to adjacent content, no epistemic status declaration, no indication of recency or verification. The reader encounters what appears to be a fact. It is actually a Particle that has been stripped of its machine-legibility governance and re-presented without it.

This is not a problem content teams can solve by optimizing for fragment extraction. It is a problem content teams can mitigate by ensuring that their content's machine-readable declarations are rich enough that extraction systems have the structural context available—even if any given AI surface chooses not to surface it. The governance obligation is to make the context available. Whether a given platform honors that context is a platform governance question outside the framework's scope, but the content system's failure to provide the context is within it.

Relationship to the Intelligence Layer and Taxonomy

The three cross-cutting dimensions are complementary, not overlapping:

1. Taxonomy governs how content is classified internally—the attributes that make the tier structure a routing and assembly system.

2. The Intelligence Layer governs how content behaves when it is dynamically generated, assembled, or delivered by AI and agentic systems.

3. Machine-Legibility Governance governs how content declares itself to the external systems that discover, retrieve, and re-present it.

A content object can be well-governed by Taxonomy (correctly classified), well-governed by the Intelligence Layer (correctly constrained for dynamic generation), and entirely ungoverned by Machine-Legibility (invisible or misrepresented to every external system that encounters it). All three dimensions are required for full governance coverage. The practical implication: structured data, schema markup, entity declarations, topical authority signals, and content freshness markers are not SEO tactics bolted onto finished content. They are Machine-Legibility Governance decisions that should be specified at the same time and by the same team making the content, taxonomy, and intelligence governance decisions they describe.

What This Layer Does Not Do

The Machine-Legibility Governance does not add SEO, GEO, or AEO as named dimensions or tiers. The Tiered Content Framework describes the content governance architecture that produces machine legibility when done well—it does not prescribe the technical implementation or measure search performance outcomes. Practitioners working in SEO, GEO, and AI Experience Optimization should recognize those concerns addressed structurally here. The framework's value to those disciplines is that it locates that work within a governance architecture rather than treating it as a post-production optimization layer.

The three cross-cutting layers, Intelligence, Taxonomy, and Machine-Legibility, together constitute the governance dimension of the Tiered Content Framework. The TCF as a whole composes with adjacent architectures as the Governance layer of the Seam Stack, a four-layer composition documented at seamstack.org:

1. Substrate — where the data lives and who owns it (Solid Pods, WebID, Access Control Policy)

2. Governance — how meaning is structured, classified, and made machine-legible (the TCF)

3. Boundary — the discipline of transitions between systems and parties (the seam-as-architectural-object)

4. Evidence — contemporaneity, tamper-evidence, and deferred-party legibility (Verifiable Credentials, RFC 3161 timestamps, OpenTimestamps)

The TCF governs meaning. It does not govern storage, ownership, transition discipline, or evidentiary durability; those are peer layers the TCF composes with, not dimensions inside it.

Applied Work & Practical Applications

The framework has been applied across enterprise, brand, and digital product engagements. Its primary contribution in practice has been providing a shared structural vocabulary that bridges content strategy, UX, design systems, CMS architecture, and engineering—reducing coordination overhead and making structural content decisions traceable, scalable, and governed rather than ad hoc.

It forms the theoretical backbone of the Content Strategy Product Suite—a modular platform that transforms content strategy from a consulting deliverable into a governed, repeatable, data-driven workflow.

Here are practical applications of the Tiered Content Framework across common enterprise and agency scenarios:

Content Audit & Inventory

Rather than auditing a site as "pages," you audit it by tier. You identify ungoverned Quarks (missing terminology rules, undefined tone parameters, absent schema constraints), orphaned Particles (fields with no governing rules downstream of them), broken Clusters (components whose semantic objects don't hold together), and missing Zones (page areas with no clear user intent). This gives you a structured gap analysis instead of a subjective quality review.

Enterprise CMS Architecture

When building or migrating a CMS, each tier maps to a content model layer. Particles become structured fields with validation rules. Clusters become content types. Zones become layout regions. Structures become templates. This gives the CMS architecture a semantic foundation—not just a presentation model.

AI Content Governance

The Intelligence Layer makes the framework directly applicable to AI-generated content. At the Particle level, you're defining the field constraints, terminology guardrails, and tone parameters that get passed into prompts. At the Cluster and Zone levels, you're governing how AI assembles dynamic content so it maintains structural coherence and intent alignment—even when no human authored it.

Brand Consolidation & Mergers

When two organizations merge digital presences, you can map each independently to the Biome/Ecosystem tiers, then identify where Structures and Zones overlap, conflict, or can be consolidated. It turns an abstract "content rationalization" project into a structured comparison with clear governance decisions at each tier.

Content Briefing Systems

Briefing templates can be built at the Cluster or Zone tier, so instead of briefing "a hero section," you're briefing a Zone with defined intent, required Clusters, and Particle-level constraints already embedded. Writers and AI systems receive structurally complete briefs, not blank slates.

Personalization Architecture

Personalization often fails because it operates at the page level, swapping whole pages rather than targeted semantic objects. The framework enables Cluster- and Zone-level personalization: you vary specific semantic objects (an author card, a trust signal, a CTA label) within a stable Structure, rather than forking entire experiences.

Design System Alignment

The framework gives content parity with the design system. Where a design system governs components at the UI level, the Tiered Content Framework governs the semantic layer beneath them. This enables true design-content co-governance, with every component getting a corresponding content object with its own rules, not just visual specs.

Governance Handoffs

When a content strategist leaves an organization (the "designer not present" constraint the framework was built around), the tier model serves as the handoff artifact. The next person inherits not just a style guide but a full operating model: what the content objects are, how they relate to one another, and the rules that govern each tier.

—

Additional Applications

Additional named applications will be added in as they’re identified or suggested.

The Creation Layer

Continue reading about how to apply this in practice, at the creation layer:
Tiered Content Framework Creation Layer →

Claude as Collaborator: How AI Governed the Framework That Governs AI Content

The Tiered Content Framework wasn't just developed for AI-era content governance; parts of it were developed with AI. Claude (Anthropic) served as a structured-thinking partner throughout the multi-year development arc that produced the TCF's current form, contributing most directly during the intensive spring 2026 refinement period that brought the framework to public release.

The collaboration was governed rather than generative. Claude did not author the TCF. The seven-tier architecture, the Intelligence Layer, the Taxonomy and Machine-Legibility dimensions, the production chain documentation, and the framework's core argument, that content governance should operate with the same structural precision as UI governance, are original to Jedi Wright. What Claude contributed was sustained analytical pressure: pressure-testing structural decisions, surfacing gaps between the framework's stated scope and its actual coverage, drafting and redrafting documentation until the voice was precise without being stiff, and functioning as a sounding board.

Specific contributions included: iterative development of the Creation Layer documentation and its AI-governed/AI-generated distinction; editorial refinement across the framework page, production chain post, and Figma talk track; cross-document consistency enforcement as the framework evolved; and the development of the tcf-particle-to-structure-creator skill—a Claude skill built to implement the TCF's Particle-to-Structure production logic in live content creation workflows.

That last item is the recursive case. A Claude skill, governed by the TCF's tier vocabulary, now operationalizes the framework in practice. The Intelligence Layer, the TCF's cross-cutting governance dimension for AI-assisted content, was shaped in part through the experience of actually using AI to build and stress-test the framework it would eventually govern.

This is not a story about AI replacing content strategy judgment. It's a case study in what AI-governed content development looks like at the framework level: human direction, human intellectual ownership, AI as a disciplined structural collaborator operating within a governed production process.

Acknowledgements

With thanks to former Hero teammates Andrew Kaufman, Brian Lynn, Doug Holton, Jenn Vitello, and others I may be forgetting…message me, and I'll add you.

Content Strategy as Structural Infrastructure: Extending Atomic Design Methodology for Governed, Scalable Digital Experiences

Jedi Wright · v0.1 · Independent research · 2021–2026

Full paper available on request. Citation and academic inquiries:
jedi@jediwright.com

Interested in applying this framework to your organization's content architecture, or want to weigh in on where it holds up and where it breaks? Work with me →

Changelog

v1.7, May 26th, 2026
Added field-level specification for the Epistemic Status schema within the Machine-Legibility Layer. v1.6 correctly established the governance architecture: four status values (confirmed / inferred / unverified / time-sensitive), governed at the Quark level, declared at the Particle level, inherited upward through the tier structure by the weakest-status rule, but stopped short of specifying the concrete fields that make the declaration implementable. A practitioner could follow the governance logic, but couldn't build from it. v1.7 closes that gap.

The five governed fields (tcf:epistemicStatus, tcf:authoritySource, tcf:confidenceLevel, tcf:temporalValidity, tcf:verificationRecord) now carry data types, required/optional conditions, and JSON-LD property mappings. The propagation rule is formally expressed, including expiry degradation: a time-sensitive Particle whose recency window has passed propagates as unverified rather than time-sensitive, because an expired verification is no longer current conditions. The tcf:aiProvenance field schema formally establishes AI generation method as a distinct attribute from epistemic verification state; the two fields govern different things, and neither determines the other.

The vocabulary alignment between the TCF's epistemic status values and my Agentic Accountability Playbook (coming soon) is formalized with a maintenance protocol: if either framework proposes a vocabulary change, a 14-day cross-framework review is required before publication, with the outcome being alignment or an explicit fork with a documented rationale, recorded in both changelogs.

Expanded the Taxonomy section with a tier-by-tier breakdown parallel to the Intelligence and Machine-Legibility layers, corrected a Quark-sync issue introduced in v1.5–1.6 (classification is governed at the Quark level and declared at the Particle level; the prior text had these reversed), and named the static boxes failure mode: the diagnostic condition a well-governed tier structure produces when taxonomy is absent.

Context Spectrum Mapping appendix—forthcoming.
A positioning appendix is planned for a future minor revision that maps TCF constructs relative to the statistical proximity → typed relations → formal ontological commitment spectrum. The framework operates across all three bands, with Quark-level schema constraints at the formal commitment end, Taxonomy in typed relations, dynamic content assembly at the statistical proximity end, but that mapping is not yet surfaced as navigable documentation. The appendix will make the implicit explicit and improve legibility for knowledge graph and semantic web practitioners building at that layer.

v1.5-1.6, May 23rd, 2026
Updated framework tier count from six to seven (Tier 0 through Tier 6) following formal specification of Tier 0, Quarks. Quarks are the raw, format-independent constraints, rules, and values that Particles are made of and governed by—not content, but the conditions under which content is valid. Examples include approved terminology lists, character limits, tone parameters, taxonomy values, metadata schema definitions, accessibility thresholds, brand voice primitives, and prompt-engineering constraints governing AI-generated content. Quarks carry the highest change-control scrutiny in the framework: every modification propagates downstream through every Particle, Cluster, Zone, Structure, Ecosystem, and Biome that inherits from it.

Updated the Intelligence Layer, Taxonomy, and Machine-Legibility sections to reflect Quarks' relationships to each cross-cutting dimension: the rules governing machine-readable declarations originate at the Quark level and are inherited by Particles; taxonomy values are Quark-level constraints that precede and govern Particle-level classification; prompt engineering constraints fed into LLM-generated Particles are governed at the Quark level. Naming rationale: "Quarks" is drawn from the same physical metaphor the framework already uses, Particles, and correctly positions this layer as sub-atomic. "Tokens," the term used in Atomic Design's equivalent layer, was evaluated and rejected to avoid terminology collision in content contexts. Consolidated cross-framework alignment language following parallel development of another framework I’m developing, called the Narrative Content Framework (NCF): the NCF inherits its tier structure, governance principle, KER logic, and cross-cutting dimension architecture from the TCF; NCF development in turn produced the structural insight that identified the Quarks amendment—the relationship is bidirectional.

Removed a precedence claim that did not survive stress testing: the distinction between the TCF and Atomic Design's equivalent layer is governance depth and specification rigor, not priority of insight. Also, corrected the Atomic Design cross-reference column in the tier table to reflect Tokens as the structurally equivalent layer, an omission in prior versions that understated the parallel and left the cross-framework alignment incomplete.

v1.4, May 4th, 2026
Updated the Claude as Collaborator section.

v1.3, April 19th, 2026
Updated tier definitions for Zones and Structures to be endpoint-agnostic, and added named deployment contexts to each. Prior definitions anchored Zones as "page-area containers" and Structures as "page-level compositions"—language that breaks down in headless, omnichannel, and AI assistant environments where the presentation layer is entirely decoupled from the page. Zones are now defined as context containers: functional regions within a Structure that govern content assembly for a specific purpose. Structures are now defined as endpoint compositions: the complete, governed assembly of Zones delivered to a specific surface. Named deployment contexts are added to each tier definition, specifying how Zones and Structures function when the endpoint is a web page, app screen, voice response, AI assistant answer, digital billboard, or watch face. The remaining tiers—Particles, Clusters, Ecosystems, and Biomes—are inherently endpoint-agnostic and require no revision. Responsive to practitioner feedback from a Director of AI Content Strategy, identifying that page-centric terminology in the upper tiers misrepresents how the framework operates in modern headless and liquid content environments, where the endpoint is a delivery target rather than a defining characteristic of the tier. No changes to the six-tier model structure or the three cross-cutting governance dimensions.

v1.2, April 16th, 2026
Added the Machine-Legibility Layer as a third cross-cutting governance dimension, peer to the Intelligence Layer and Taxonomy. Governs how content at each tier declares its identity, relationships, authority, and epistemic status to search engines, knowledge graphs, AI retrieval systems, and large language models. Names the AI Search Fragment Problem as the dimension's diagnostic failure mode: a Particle-level extraction stripped of the governance context that gave the original content its meaning. Responsive to practitioner feedback from an SEO and GEO practitioner, identifying that the Intelligence Layer and Taxonomy together address routing and assembly logic, but leave ungoverned the technical surface through which AI systems understand content relationships—structured data, schema, entity relationships, topical authority signals, and internal linking architecture. An ECD practitioner thread sharpened the specific failure condition: no-click AI search surfaces content fragments without the metadata, schema, and relational context required for accurate retrieval, confirming that the content governance framework must mirror the code content structure at the Particle level. The dimension extends the framework's reach from content strategy and governance into the technical SEO and GEO layer, which practitioners identified as going hand in hand rather than being separable concerns. No changes to the six-tier model or the two existing cross-cutting dimensions.

v1.1, April 14th, 2026
Added Taxonomy as a second cross-cutting governance dimension, peer to The Intelligence Layer. Governs machine-readable classification across every tier, providing the routing, assembly, and personalization logic that makes the tier structure machine-actionable. Classification originates at the Particle level—structured fields carrying attributes such as Content_Type, Audience, and Intent—and cascades upward through Clusters, Zones, and Structures, where dependencies and zone-affinity rules are validated. Names the static boxes’ failure mode as the dimension's diagnostic condition: tiers without taxonomy remain a governance vocabulary rather than a dynamic, flowing system. Responsive to practitioner feedback, identifying that the Intelligence Layer's references to structured fields, semantic tagging, and dynamic assembly gestured at taxonomy without naming or operationalizing it as a formal construct; insufficient for practitioners building personalization systems or AI-driven experiences. Taxonomy is not a seventh tier; it is a governance dimension that cascades through every tier rather than residing at one. No changes to the six-tier model or The Intelligence Layer.

v1.0, April 13th, 2026
Initial publication. A six-tier content governance model that extends Brad Frost's Atomic Design methodology to content strategy and information architecture. One cross-cutting governance dimension: The Intelligence Layer, governing how each tier behaves when content is dynamically generated, assembled, or delivered by AI and agentic systems. Creation Layer production chain documented. The Content Strategy Product Suite is named as the commercial implementation.