AI is changing how people find and interact with government information. Increasingly, users may not begin their journey on a government website at all. They may ask a search engine, an AI assistant or an emerging digital service a question and expect a clear, accurate answer.
For government, this creates an important challenge:
How do we ensure those answers are based on authoritative, current and trusted government information?
The answer isn't simply to add AI to existing websites.
Government needs to ensure the information underneath these services is trustworthy, well governed, structured and able to be understood by both people and machines.
The good news is that government doesn't need to start from scratch. Work already underway through the NSW Government OneCX Program and the Australian Government's GovCMS Content as an API (CA3) initiative provides useful insight into what AI-ready government information infrastructure could look like.
Together, these initiatives also provide practical lessons that government organisations can start applying now.
Start with trusted information
Before thinking about AI models, agents or chatbots, there is a more fundamental question:
Can we trust the information they will use?
An AI system can retrieve, summarise and combine information at extraordinary speed. But if the underlying information is duplicated, contradictory, outdated or poorly structured, AI can amplify those problems just as quickly.
This makes information governance an important part of AI readiness.
The NSW Government OneCX Program provides a useful example of this principle in practice.
OneCX is an all of NSW government initiative migrating 450+ government agency websites onto nsw.gov.au, with information organised around customer needs rather than the structures of government.
As part of this process, content can be reviewed, improved and consolidated into a more consistent information environment.
The OneCX team's guidance on how to optimise content for AI search highlights several foundations that are increasingly important in an AI-enabled environment.
These include removing duplicate or conflicting content during migration, organising information into customer-centric topics, using consistent content structures, rewriting content into plain language and providing tools to help content teams identify quality issues.
These practices benefit people using government websites today, while also making government information easier for search engines and AI systems to understand.
AI is already changing government search
This isn't only preparation for a future technology.
NSW Government has already introduced AI-generated search summaries on nsw.gov.au.
Rather than requiring users to navigate a conventional list of search results, generative AI provides a short summary at the top of relevant searches to help people find the information they need more quickly.
Importantly, the summaries are grounded in NSW Government information. They draw from the top relevant search results across nsw.gov.au and service.nsw.gov.au, with links to the underlying pages available for further information.
Before launch, the OneCX team tested generated responses against criteria including accuracy, freshness, fairness and bias, and inclusiveness. The feature was also reviewed against the NSW AI Assessment Framework.
This demonstrates an important change in the role of government search.
Traditional search asks:
Which pages contain the words the user entered?
AI-enabled and semantic approaches increasingly allow us to ask:
What is this person trying to understand, and which authoritative information can answer their question?
That can create significantly better experiences for citizens, but it also increases the importance of the information sitting behind the answer.
If government is going to generate answers, it needs confidence in the sources from which those answers are generated.
CA3: looking across the wider government information estate
The Australian Government's GovCMS team is exploring another part of this challenge through Content as an API (CA3).
CA3 is progressing from Alpha towards Beta, with GovCMS developing a Content Corpus and capabilities designed to help government organisations get greater value from their existing content.
Five emerging CA3 capabilities are particularly interesting.
Find: semantic discovery across government content, moving beyond traditional keyword search.
Verify: support for checking AI-generated responses against authoritative government content.
Align: identification of potentially duplicated, conflicting or overlapping content.
Migrate: support for moving existing content towards more structured and AI-enabled architectures.
Weave: analytics and insight across government content to help understand the wider information estate.
Taken together, these capabilities point towards something larger than an improved search engine.
They suggest an emerging information layer through which government content can be discovered, compared, analysed, verified and improved.
OneCX and CA3: two complementary strategies
There is significant alignment between the OneCX and CA3 approaches, but they operate at different layers of the problem.
OneCX demonstrates the value of consolidating and governing authoritative information at source.
CA3 explores what becomes possible when government can understand and interrogate authoritative information across a wider, distributed information estate.
Their approaches can be considered side by side:
| CA3 capability | Alignment with OneCX |
|---|---|
| Find: semantic discovery across government content, moving beyond traditional keyword search. | AI-enabled search and content optimised around customer intent and discoverability |
| Verify: support for checking AI-generated responses against authoritative government content. | Ensuring AI-generated information is grounded in authoritative government sources |
| Align: identification of potentially duplicated, conflicting or overlapping content. | Identifying and reducing duplicate, overlapping and conflicting content |
| Migrate: support for moving existing content towards more structured and AI-enabled architectures. | Moving legacy information into better structured, governed and reusable content |
| Weave: analytics and insight across government content to help understand the wider information estate. | Using analytics and insight to understand content performance and information needs |
The technologies and scope are different, but the underlying principle is remarkably consistent:
Better AI starts with better information.
One way to think about the distinction is:
OneCX: consolidate the experience and improve the information at source.
CA3: create an intelligence layer across distributed authoritative information.
Government will increasingly need both approaches.
From finding information to understanding the information estate
One of the most interesting opportunities is the shift from simply retrieving government information to understanding the relationships within it.
Imagine three government sources providing slightly different guidance about the same service.
Traditional search may successfully find all three pages.
Semantic search may recognise that all three relate to the same citizen need.
But the next generation of government information infrastructure can potentially ask a more important question:
Why is government giving different answers to the same question?
This is where capabilities such as CA3's Align become particularly interesting.
AI and semantic technologies can potentially help identify duplication, inconsistency and conflicting advice across large information estates that would be extremely difficult for people to review manually.
AI then becomes more than a mechanism for generating answers.
It can become a tool for knowledge governance.
That has implications well beyond search. It could help government better understand what information it holds, where information overlaps, where potential conflicts exist and where stronger governance is required.
What can government organisations do now?
You don't need to wait for these initiatives to mature before beginning this work.
Many of the foundations required for AI-enabled services are also good digital and content practices today.
Government organisations can start by:
- Identify authoritative sources. Know which information represents the source of truth, who owns it and who is accountable for keeping it current.
- Find duplication and conflicting information. Understand where multiple pages, websites or documents provide overlapping or inconsistent advice.
- Structure content around meaning. Move beyond pages and documents where appropriate towards reusable, structured content that describes what the information actually represents.
- Improve taxonomy and metadata. Consistent classification helps people, search technologies and AI systems understand the meaning of information and relationships between it.
- Establish provenance. Make it possible to understand where information came from, who owns it, when it changed and whether it remains authoritative.
- Reduce information trapped in documents. Important information contained only in PDFs and other documents can be more difficult to discover, reuse and interpret. Consider where structured web content is more appropriate.
- Make information reusable. APIs and structured interfaces allow authoritative information to support websites, search, applications, AI services and channels that may not yet exist.
- Design governance for people and machines. Content governance increasingly needs to consider not only what a person sees on a webpage, but how machines may retrieve, interpret and reuse that information.
These are useful investments regardless of which AI technologies ultimately become dominant.
The role of open platforms
This emerging environment also changes how we should think about content management systems.
Historically, the primary publishing model was relatively straightforward:
Person > browser > webpage
Government information increasingly needs to support a much broader ecosystem:
People > search > APIs > applications > AI assistants > AI agents > future digital services
The underlying information therefore needs to remain structured, governed, portable and accessible.
This is one of the areas where open source platforms such as Drupal have an important role to play.
Drupal has long treated content as structured information rather than simply webpages. Content types, fields, taxonomy, metadata, revision history, workflows, permissions and APIs provide foundations for information that can be governed and reused across multiple channels.
This doesn't mean Drupal, or any CMS, is itself an AI strategy.
It means that open, structured and extensible platforms can provide part of the information infrastructure upon which AI-enabled services depend.
There is already a strong Australian government precedent. Drupal underpins both the NSW Government's OneCX platform and the Australian Government's GovCMS platform and is used extensively across Commonwealth, state and local government.
The opportunity now is to build on that investment.
Before buying an AI solution, ask these questions
For government organisations considering AI-enabled digital services, a useful starting point may have very little to do with selecting an AI product.
Ask:
Can we identify our authoritative information?
Do we know who owns it and is accountable for its accuracy?
Is important information structured and machine-readable?
Can we identify duplicated or conflicting advice?
Do we use consistent taxonomy and metadata?
Can authoritative information be accessed through APIs or structured interfaces?
Can we trace information back to its source and understand when it changed?
Can an AI-generated answer be verified against authoritative government information?
If the answer to several of these questions is no, that is a useful indication of where to start.
Design government information for people and machines
The next generation of digital government won't simply be about putting AI interfaces in front of existing websites.
It will require government to think differently about the information underneath those experiences.
The work underway through NSW OneCX and GovCMS CA3 provides two useful views of that future.
OneCX demonstrates what can be achieved by consolidating, improving and governing authoritative government content.
CA3 explores what becomes possible when technology can discover, verify, compare, migrate and analyse authoritative information across a much broader government information estate.
Both point towards the same conclusion:
Trusted information is the foundation of AI-enabled government.
The organisations best prepared for AI may therefore not be those that adopt AI first. They may be those that first ensure their information can be trusted, understood, governed and reused by people and machines alike.
Find out more
These government initiatives provide useful resources for digital, content and technology teams considering where to start.
NSW Government OneCX
How to optimise your content for AI search
OneCX guidance on authoritative, structured and consistent content and how established content practices can improve visibility and interpretation by AI systems.
nsw.gov.au launches AI search summaries
A practical example of generative AI being introduced into an established government digital platform, including how summaries are grounded in government sources and tested for responsible use.
About the OneCX Program
More information about the all-of-government program bringing NSW Government websites and information together on nsw.gov.au.
Australian Government GovCMS
Content as an API (CA3): Progress towards Beta
The latest GovCMS update on CA3 and its FIND, VERIFY, ALIGN, MIGRATE and WEAVE capabilities.
Where to start
You don't need to wait for the next generation of AI technology to prepare for AI-enabled government services.
Start with the information you already have.
Identify your authoritative sources. Reduce duplication and conflicting advice. Improve content structure, taxonomy and metadata. Establish ownership and provenance. Make important information machine-readable and reusable.
Most importantly, ask whether an AI system could find the right information, understand it and trace an answer back to an authoritative government source.
These are good digital practices today; and increasingly important foundations for the AI-enabled government services of tomorrow.
At the Government Innovation Showcase NSW?
Visit the Drupal stand to talk with members of the Australian Drupal community about structured content, AI readiness, open source and the role trusted information can play in the next generation of government digital services.