From Blobs to Bots: How Structured Content Predicted Machine Experience

6 March 2026 · 19 min read

In 2017, Carrie Hane and Mike Atherton published "Designing Connected Content: Plan and Model Digital Products for Today and Tomorrow", a guide to creating content that works across any channel or device. Their central insight—that content should be structured as discrete, meaningful chunks rather than monolithic blobs—was aimed at solving a human problem: how to publish the same content efficiently across websites, mobile apps, voice interfaces, and whatever comes next.

But in making content "future-friendly" for humans across multiple channels, Hane and Atherton inadvertently created the blueprint for Machine Experience. Because content structured for flexibility and reuse is, by definition, content structured for machine understanding.

The content strategists were building for AI before AI was the obvious use case.

I'd arrived at the same conclusion independently before I encountered Hane and Atherton's work. In early 2024, a piece I wrote for CMS Critic after CMS Kickoff argued that AI's real strength isn't creating content — it's consuming it. But most websites bury their meaning in visual layout that machines can't parse. I compared how an AI reads a web page to how an eight-year-old shops for toys: ignoring brand loyalty, navigation hierarchies, and carefully designed user journeys. It just wants the answer. That article was the start of Machine Experience as an idea. When I later discovered "Designing Connected Content," I realised the structured content community had been building toward the same insight for years.

Blobs Versus Chunks: A Tale of Two Architectures

Hane describes the traditional approach to web content as "blobs"—everything dumped into a single WYSIWYG body field. A typical blog post might include the title, author bio, publication date, body text, images, captions, and related links all mashed together in one undifferentiated mass.

For humans viewing a designed page, this works. The visual presentation implies relationships: the byline sits under the headline, the image caption appears below the image, the author bio is in a sidebar. Humans infer structure from layout.

But try to reuse that content elsewhere—say, in a mobile app or voice interface—and you're stuck. The structure exists only in the visual presentation, not in the content itself. Want just the headline for a notification? It's buried in the blob. Need the publication date for sorting? It's mixed in with everything else. Want to pull all articles by a specific author? Good luck parsing that.

This is where Hane and Atherton's "chunks" come in. Instead of one blob, content is broken into discrete, meaningful pieces: headline (text field), author (relationship to person entity), publication date (date field), body paragraphs (structured text), images (media objects with their own metadata), and so on.

Each chunk is defined, typed, and explicitly related to other chunks. As Hane notes, "the chunks of content became data, with the relationships explicitly described in a way that both people and robots can understand."

Notice that phrase: "both people and robots."

Domain Modeling: Teaching Machines the Truth

The structured content methodology begins with domain modeling—mapping out the real-world concepts and relationships in your subject area before considering how they'll be presented.

If you're building a conference website, your domain model might include concepts like Sessions, Speakers, Venues, Sponsors, and Time Slots. Before you design a single page, you map out how these concepts relate: a Session has one or more Speakers, takes place in a Venue, occupies a Time Slot, and might have zero or more Sponsors.

This isn't about web design—it's about modelling reality. You're creating what Hane calls "the truth of the domain."

For Machine Experience, this matters. When you explicitly model domain concepts and relationships, you're creating exactly what AI systems need: a semantic map of meaning. You're not asking the AI to infer that "Jane Smith" appearing under "Speaker" probably means Jane is speaking at this session. You're declaring the relationship explicitly in your content model.

Domain modeling for humans translates directly to semantic clarity for machines.

Content Types: Structured Data by Another Name

Once you have a domain model, Hane and Atherton advocate creating "content types"—templates that define what information a piece of content contains and how it's structured. A Session content type might include fields for title, description, speaker(s), time slot, venue, and sponsors.

Sound familiar? This is structured data. A Session content type is functionally identical to schema.org's Event type. Both define a template for representing a real-world concept with specific properties and relationships.

The difference is audience. Hane and Atherton were thinking about content management efficiency and multi-channel publishing. Schema.org was thinking about search engines and knowledge graphs. But they converged on the same solution: explicit, structured representation of concepts and relationships.

When you build content types following structured content principles, you're simultaneously building machine-readable structured data. The methodologies are identical because the underlying problem is the same: how do you represent meaning explicitly rather than implicitly?

COPE: Create Once, Publish Everywhere—Including to AI

A core principle of structured content is COPE: Create Once, Publish Everywhere. Write your content as structured chunks, then assemble and present those chunks differently for different contexts—website, mobile app, email, print, voice interface.

The BBC famously adopted this approach, creating a structure that supports 1,500 new shows being added every day. Each show is defined once as structured data, then presented across dozens of different interfaces and experiences.

For Machine Experience, COPE extends naturally to: Create Once, Publish to Humans and Machines. The same structured content that enables efficient multi-channel human publishing also enables AI systems to understand, process, and reuse that content.

When content is properly chunked and structured, an AI can:

Extract specific facts without parsing visual layout
Understand relationships between concepts
Reuse content in different contexts
Aggregate information across many sources
Generate accurate summaries and responses

The content strategy community solved this problem years ago. They just solved it for human channels first.

Metadata: Making Meaning Explicit

Hane emphasises applying metadata to chunks to "embed meaning so that computers can understand what it is, what it is about, and how it relates to other chunks."

This is the essence of Machine Experience. Don't rely on context, position, or visual styling to convey meaning. Make it explicit through metadata and structured relationships.

A byline that appears visually under a headline might be understood by humans as "this person wrote this article." But to a machine, it's just text in a certain position unless you explicitly mark it up with authorship metadata or a relationship to a Person entity.

Structured content advocates learned this lesson in the context of content management systems and multi-channel publishing. MX applies the same principle to machine readability: explicit is better than implicit. Always.

Future-Friendly Content Is Machine-Friendly Content

The phrase "future-friendly" appears throughout Hane and Atherton's work. The idea is that you can't predict what devices, channels, or interfaces will exist in five years, but you can structure your content so it's ready for them.

How? By separating content from presentation. By chunking content into meaningful units. By explicitly declaring relationships. By modelling concepts and meaning rather than just visual layout.

These principles create future-friendly content because they create flexible content—content that doesn't depend on a specific presentation context to be understood.

But flexibility for unpredictable future human interfaces is the same as readability for current AI systems. Both require structure. Both require explicit relationships. Both require separation of content and presentation.

AI systems are, in a sense, the ultimate future interface that structured content was preparing for. They're channel-agnostic by nature. They process meaning, not layout. They need explicit relationships, not visual inference.

Content structured according to Hane and Atherton's principles is inherently AI-ready, even though AI wasn't the primary consideration when those principles were developed.

The Implicit Becomes Explicit

Hane notes that in traditional web content, "relationships between the speakers, their sessions, and the session's time, duration, and location are only implied on this one page." Visual layout creates the impression of relationship, but the structure isn't actually there.

In structured content, those relationships are explicit. A Session entity has a declared relationship to Speaker entities, Venue entities, and TimeSlot entities. The relationships exist at the data level, not just the presentation level.

For humans, this enables efficient content reuse across channels. For machines, this enables accurate understanding without inference.

This is the core parallel between structured content and Machine Experience: both recognise that implicit relationships (conveyed through position, styling, or context) are fragile and context-dependent. Explicit relationships (declared in data structure) are robust and context-independent.

Chunks Are Parseable Units

The "chunk" concept in structured content maps directly to processing efficiency in AI systems. When content is broken into well-defined, typed units, each chunk can be processed independently and recombined as needed.

For content strategists, this means you can pull the headline for a notification, the excerpt for an email, the full text for a webpage, and the audio version for a podcast—all from the same structured source.

For AI systems, it means you can process discrete units of meaning without having to parse and separate them from surrounding content first. The chunking is already done. The boundaries are explicit. The types are declared.

Just as chunked content makes multi-channel publishing efficient, it makes AI processing efficient. Same principle, different application.

Content Is Data

Hane emphasises that "content is data". It sounds obvious, but it changes how you think. When you treat content as data, you think about structure, types, relationships, and schemas. You think about how pieces fit together, not just how they look.

This is exactly the mindset shift required for Machine Experience. Stop thinking about content as text on a page. Think about it as structured data that can be queried, analysed, and processed.

The content strategy community made this shift to enable better content management and multi-channel publishing. The same shift enables better machine readability.

The Five-Step Framework Maps to MX

Hane and Atherton's framework for structured content includes:

Domain modeling — Model the truth of your subject area
Content modeling — Define content types and their properties
Content design — Design content based on structure
CMS implementation — Build the content repository
Interface design — Design templates and navigation

For Machine Experience, we can map this almost directly:

Semantic modeling — Define concepts and relationships
Data schema definition — Use schema.org or custom schemas
Semantic markup — Structure content with meaningful HTML
Content architecture — Organise for both storage and retrieval
Presentation layer — Design interfaces (human and machine)

The methodologies parallel because the underlying challenges parallel: how do you represent complex information in ways that can be understood, reused, and presented flexibly?

The Convergence

What's striking about structured content and Machine Experience is how they arrived at similar solutions from different starting points.

Structured content began with the problem of content proliferation across channels. How do you manage thousands of pieces of content that need to appear on websites, apps, emails, print materials, and voice interfaces—without creating separate versions for each?

Machine Experience begins with the problem of AI comprehension. How do you structure content so machine systems can understand, process, and use it accurately?

The solutions converge because both problems ultimately require the same thing: explicit structure, declared relationships, separation of content and presentation, and semantic clarity.

Hane and Atherton were optimising for content management efficiency and multi-channel publishing. But in doing so, they created the architecture AI systems need.

What Content Strategists Can Teach MX Practitioners

The structured content community has years of experience with the practical challenges of implementing these principles:

Hane addresses getting organisational buy-in in her "Convincing Your Boss" chapter. The same arguments work for MX: better SEO, improved findability, reduced duplication, future-proofing.

You don't have to restructure everything at once, either. Start with new content types. Model content without immediate CMS changes. Work one section at a time.

Domain modelling requires subject matter experts. Content modelling requires collaboration between strategists, designers, and developers. The same cross-functional approach works for MX. And the structured content community has already worked out how to balance structure and flexibility — too much becomes rigid, too little becomes chaos.

MX practitioners can learn from these proven approaches rather than reinventing them.

The Content Strategy Precedent

Perhaps the most valuable insight from the structured content movement is simply this: it proves the approach works.

The BBC, NPR, and other organisations have demonstrated that content structured this way is manageable, scalable, and future-friendly. It works across channels. It enables personalisation. It reduces duplication. It makes content findable.

All of these benefits exist before you consider AI. They're valuable for purely human content management and publishing.

But the structured content that enables these benefits is also, inherently, machine-readable. The architecture that makes content future-friendly for unpredictable human interfaces also makes it comprehensible to current AI systems.

This isn't accidental. Explicit structure, declared relationships, and semantic clarity are fundamental principles of information architecture — whether you're building for content managers, mobile devices, or language models.

The Unified Principle

Carrie Hane and Mike Atherton's work on structured content and the emerging discipline of Machine Experience share a unified principle: meaning should be explicit, not implicit.

For structured content, this enables efficient content management and multi-channel publishing.

For Machine Experience, this enables accurate AI understanding and processing.

But at a deeper level, both are simply arguing for better information architecture. Content should be structured according to what it means, not just how it looks. Relationships should be declared, not inferred. Concepts should be modelled, not assumed.

These principles create better content for humans and machines. They enable both multi-channel publishing and AI processing. They're future-friendly and machine-friendly because, ultimately, they're just well-designed information.

The content strategists figured this out first. Machine Experience is, in many ways, just extending their work to a new audience.

A note on names: The MX community is called The Gathering — a name drawn from the Scottish clan tradition, where clans gather to decide the best future for the combined assembly, then go their separate ways. This predates any connection to Carrie Hane's work or her company GatherContent. The similarity is a coincidence, not an influence.

Key Principles from Designing Connected Content

Structure as Liberation: "Structure creates more flexibility than amorphous blobs of text that stand alone unless manually linked together."

Content is Data: "Content is data. A piece of information does not need to be recreated for every channel and interface and delivery method."

Explicit Relationships: "The chunks of content became data, with the relationships explicitly described in a way that both people and robots can understand."

Future-Friendly: "Content structured properly can be reusable. You can turn a webinar into a blog post, 3 five-minute videos, and a downloadable slide deck."

Domain Truth: "Domain modeling is modeling the truth of the domain and the subject area that you're working in."