Posted: 30th March 2026 by McGee Noble
Over the last few years we have been grappling with how to make sense of the government websites we work with that are part of behemoth content ecosystems.
Government sites are typically megasites, with thousands or even tens of thousands of pages. They come with wayfinding and findability challenges that can be hard to solve.
But for many governments, these mega sites also now often sit within a mega digital ecosystem. This is an ecosystem of dozens, or hundreds of websites. Separate sites for services. A semi-public intranet for transparency. Microsites for campaigns. Sites for initiatives. Some agencies have their own websites; some don’t. These kinds of mega digital ecosystems have even more wayfinding and findability challenges than a single megasite, no matter how big.
And now, adding even more layers to an already complex sensemaking problem, AI is fundamentally changing how we have to think about content. It’s doing this in (at least) three ways:
- people are using AI tools for content creation and other processes
- websites now include AI tools for users
- the way people search for information is completely different to just a couple years ago
You have likely already seen research about how much information seeking behaviours are changing, like this finding from a year ago showing 80% of consumers rely on AI-written results,. We can’t ignore the importance of optimising content so that AI information retrieval is as accurate and useful as possible.
Directing AI to real-world relevance
While artificial intelligence is (usually) pretty good at scouring content and parsing it, mistakes happen. These are often down to errors in inference – where AI training generates erroneous predictions, biases, or poor decisions. Most importantly in our world, these errors can be due to source information being unstructured, repetitive, self-contradictory, obsolete or unclear.
When AI approaches a massive dataset, it might easily choose material that sits within the same digital ecosystem but doesn’t come from the most appropriate sources or webpages. The language used might yield matches, but meaningful connections are lost.
Why content strategists need to deal in meaning
This is why we are now moving towards a shift in our content types and structures that begin with an ontological model of real-world meaning. Because this is what makes content ready for AI discoverability and accurate interpretation.
Let me give you an example of what I mean:
This is what a list of content types looks like in a usual government website.
Content type list:
- Standard page (90% of pages use this)
- Home page
- Banner module
- Step by step page
Notice how they don’t communicate anything about the actual meaning of the content? And notice how almost all of the content ends up as ‘standard pages’?
When you have 10,000 pages sitting in something called a ‘standard page’ you are basically dealing with each one individually. Every page is a snowflake, because when you only have one template you really have no template.
For example, here are some content types from an ontological information model we developed for a state government client:
- Rules and laws
- Service summary
- Help entry
We articulated these content types in a model that showed the relationships between them. For example, a service summary can be related to multiple rules and laws, and to multiple help entries.
This type of ontology crystallises the relationships, constraints and interactions between things. It does more than inform content strategy. It wraps information around meaning. And that allows us to also establish governance processes that sit behind the content curtain, ensuring persistence and clarity of that meaning.
We can then also:
- ensure AI retrieval is using content that is founded on our own definitions of meaning, grounding it in context
- create highly structured content briefs that can be used alongside AI tools to accelerate large scale content improvements
Across the project we came to realise the importance of not just structured content, which is content that is structured for re-use, but of a content system that is fully structured around meaning.
Getting to these models is critical foundational work. And they do more than act as content strategy.
We summarised our work in a presentation which we shared at the DrupalSouth Community Day in Canberra in 2025 and will deliver at the We are content strategy and design meetup later this year. Be sure to join the group to get notified when it’s announced.
More if you’re interested:
Angus has talked about encapsulation at our meetup – this concept really helped us develop our thinking on content systems and meaning.
The most recent AI article to capture our attention was Claude Cowork gave me AI-induced mania by Kate Moran. Notice that the utility of the tool for her required it to be grounded in an ontological model of her own work!
Also, a fantastically detailed breakdown about ontology grounded retrieval in AI.
