Share This

Opinion

What does 'AI-native CMS' actually mean? A buyer's guide for 2026

11 November 2026 — Hamilton Jones
What does 'AI-native CMS' actually mean? A buyer's guide for 2026

We are increasingly asked the same question by clients and prospects: which AI-native CMS should we be using? It has made me think it is a good time to share what we say when we are asked. 

The short answer is that "AI-native" has become a marketing claim rather than an architectural one. Almost every content management system vendor now uses it, and it has no agreed definition. The longer answer is below: how we assess the market, where sixteen leading platforms currently sit across AI capability, enterprise fit and Australian data sovereignty, and what that means if you are choosing a platform or a partner to design and build one. 

A seven-level scale for AI in content management

Before any vendor conversation, we agree what "AI-native" actually means. We use a seven-level scale that runs from no AI at all to a platform that runs its own model with no external calls. 

A seven-level scale for AI in content management

AI maturity scale for content management · 7 levels 

The value of a scale like this is that it moves the conversation from marketing language to architecture. A vendor can say "AI-native" in a headline. It is much harder to claim level 6 or 7 when asked how the content model was designed and where inference actually happens. 

What the market looks like in 2026 

The market has moved quickly, from AI plugins, to AI embedded in editorial tools, to agentic operations. Plotted honestly against the scale, it still does not look the way vendor websites suggest. 

Level 7 is empty. There is currently no buyable CMS that runs its own native large language model without calling an external service. Treat level 7 as a direction of travel, not a product you can procure. Any vendor claiming it should be asked to show you where the model runs. 

Level 6 is either a startup or a build. The platforms genuinely designed around AI from the ground up are either very young companies, which carry real continuity risk for an enterprise or government buyer, or a composable stack you assemble yourself. A headless CMS such as Sanity, paired with a cloud AI service and a content model architected for AI from day one, can reach level 6. It is reached by design and delivery, not bought off the shelf. 

Level 5 is where the mature enterprise choice lives. Optimizely, Sitecore, Adobe Experience Manager, Contentstack, Kontent.ai, Contentful and HubSpot all now offer some form of agentic capability. They are capable, well funded and enterprise-ready. They differ less on AI depth than vendors would like you to believe, and much more on where your data and prompts are processed. 

Level 4 is where the sovereign options sit. Platforms such as Squiz and self-hosted Umbraco have AI embedded in the editorial workflow and a strong story on Australian hosting and governance. They are not architecturally AI-first, but they can reach agentic behaviour when paired with the right orchestration layer. 

Most organisations currently sit at level 2: a traditional CMS, often WordPress, with AI plugins bolted on outside the content model. That gap is usually where our conversations with clients begin. 

Sovereignty is decided at the inference layer 

For Australian organisations in government, health, financial services or defence-adjacent sectors, this is the dimension that changes the shortlist. 

Data residency is not the same as data sovereignty. Storing your content in an Australian data centre does not mean the AI processes it here. When an editor asks an AI assistant to rewrite a page, the prompt and the content travel to wherever the model runs. For many SaaS platforms, that is the United States or Europe. 

Our advice is to treat sovereignty as an inference-layer decision. Both major clouds now offer Australian-region AI inference: Azure OpenAI in Australia East, and AWS Bedrock in Sydney. If your organisation already runs on Microsoft or AWS, you may be able to keep AI processing in-country using infrastructure you already govern. 

There is a trade-off we always name. New frontier models often launch on global routing first and arrive in Australian regions later. The mature approach is tiered: sovereign inference for sensitive content, global models for low-risk tasks, and a documented rule for which is which. 

How to assess an AI-native CMS, and a partner to build it 

These are the questions we put to every platform, and the ones we would expect you to put to any partner: 

Assessment questions for platforms and partners

Assessment questions for platforms and partners · 5 

The right partner should be able to answer these across several platforms, not just the one they resell. Independence matters here, because the best answer for many organisations is a combination: a proven CMS, a sovereign AI layer, and a content model designed for both people and machines. 

The answer is assembled, not bought 

Our position is that true AI-native content management is not yet something you can procure. It is something you design. The organisations that will get there first are not waiting for a vendor to ship level 7. They are building the content model, governance and sovereign AI layer now, on platforms they already trust. 

Deepend designs and builds across Squiz, Umbraco, Sanity, Webflow and HubSpot, with no commercial ties to any vendor and deep experience in regulated industries. 

So here is the question we would ask of your current platform: if your AI assistant rewrote a sensitive page today, could you say exactly where that content went? If the answer is no, talk to us. We will plot your platform on the scale, as we do for our clients, and map the shortest path up to it. 

Hamilton Jones

Hamilton Jones

Head of Strategy (AI & Transformation)

If you’d like to continue the discussion

or have something to add, feel free to reach out.

Contact Hamilton