The Website Builder Is Dead, Long Live Content Management

I can’t help but feel that many CMSs are treating AI like kids in a candy store. There are plenty of exciting things we can do with AI. But as agents, prompts, and AI features pile up, I wonder whether we are losing sight of the one thing a CMS is supposed to be good at: managing content.

The obvious part is that content has moved from being an expensive source material to a cheap commodity. Yes, it makes it possible to flood the internet with even more content nobody reads. But AI can also personalize and customize content, making it easier for users to find the information they need without having to read everything.

The less obvious part is what this means for the people buying and building digital products. The rapid development of AI is changing the way people relate to software, and it is making some customers question where they should invest at all.

I saw the first documentation of this at CMS Kickoff 26, back in January. CEO and agency founder Karla Santi presented answers to a questionnaire sent to fellow web agencies. They’re seeing customers struggling to plan digital projects, and it shows up in the sales cycle:

• Longer decision cycles
• More stakeholders involved
• More internal client hesitation
• Larger projects being broken into smaller chunks

TYPO3 agency b13 has published an account of Karla’s presentation in their blog. It shows that these observations from the US agency market resonate with experiences in Europe.

My interpretation is that clients see the rapidly developing world of AI as a risk to any procurement decision. How do they know that today’s choice is still right next month? What if something new comes around that changes everything?

Of course, initiating a website project and choosing a CMS does not lock you into yesterday’s AI. But clients are mixing up all the new things you can do with AI with the long-term reasons why you need a website — and, even more importantly, why you need a content management system.

And it is true: For some websites, you don’t need a CMS. This fact hasn’t been high on TYPO3’s agenda. Our professional and enterprise focus makes this risk less pronounced, but it does not make us immune. Customers with smaller websites will increasingly question whether they need help from professional partners.

For other CMS platforms, this has been an existential question for a long time.

After website builder platforms like Wix and Squarespace took off more than ten years ago, WordPress developers told me the bottom had fallen out of the market. Customers weren’t asking for help to build the smallest websites anymore because they could do it themselves.

I suspect Wix and Squarespace are having their WordPress moment today. The market segment they focus on can now build its sites with AI.

The real reason behind the rise of website builders is that quite a lot of websites don’t actually need content management. They are simply too small. So perhaps what AI is really killing is not the CMS. It is the need for a traditional website builder.

The CMS’s focus on websites is historically explainable. Content management grew with the rise of the personal computer in the 1980s, and dedicated content management systems took off alongside the World Wide Web in the second half of the 1990s.

Websites were a safe and good neighborhood for the CMS to grow up in. But it wasn’t very diverse. The CMS industry has spent decades looking at content through the lens of the website, and it is time to look beyond it. We still tend to think of content in terms of pages, websites, and publishing. But content has been more than websites for a long time.

A website is one destination for content. It is not necessarily the reason the content exists. That distinction becomes more important as AI changes how content is created and consumed.

Many CMSs also seem to look to the past — the very distant past, technologically speaking. The AI prompt is the reincarnation of the terminal prompt. Though familiar to developers and quick to implement, it is far from the user experience most content editors are used to. A prompt can hide what options are available, as well as the content structure and context you are working in.

I have seen many AI demos across a wide range of CMSs. Most don’t even show the CMS UI. The work goes on within the external, prompt-based interface of an AI integrated headlessly with the CMS. This is akin to a magical trick. You enter a command, look at a progress animation, and something must have happened, because your website has updated.

AI prompts are useful for exploring ideas and generating content. But if you don’t need the CMS UI and you only work in a prompt, do you need the CMS in the first place? You’re building content, not managing it.

The UI of a CMS is not just a place to click buttons because humans are used to graphical interfaces. It makes the context visible. It shows structure, relationships, permissions, workflows, history, and status. It lets people understand what they are managing. That becomes more important — not less — when AI can make changes at a scale that no human could manage manually.

Content is more than websites, and it has been true for a long time. CMSs aren’t there to build websites. They’re there to manage content. Content for any platform. (All the way from traditional passenger information systems or museum displays to the yet-to-be-imagined who-knows-what.)

A CMS is where we decide what content exists, how it is structured, who can change it, who approves it, where it is used, and when it should be published, updated, or removed. The last part is easy to overlook when content creation becomes cheap.

Managing content is a different matter. The hard part is no longer creating the page. It is managing everything around it. Even if AI makes content production easier, it does not make it easier to decide what should be produced, who may publish it, or whether it can be trusted. 

A global organization might need translation into ten languages, review by regional teams, approval by marketing and legal, and publication in several channels. Six months later, every occurrence may need to be found, reviewed, and updated, across languages, teams, and channels.

In fact, when content becomes cheap, context and provenance become more valuable. Knowing who created or changed something, whether AI was involved, who approved it, which version is authoritative, and where it is being used is part of content management.

When content is generated or changed by AI, that information is not only useful for understanding where it came from. It can also help establish responsibility for what was ultimately published.

What really died with the advent of AI may be the website builder. What remains is the need to manage content effectively. This is also where structured content matters.

TYPO3 has always been strong at separating content from presentation and organizing structured content. That is not made obsolete by AI. It is time to take that strength beyond the assumption that content is ultimately destined for a website. Structured content becomes even more valuable when content needs to carry information about its provenance, approval, and where and how it is published.

I believe three things are needed for the content management system of the future. These are not implementation proposals or a roadmap for TYPO3, but areas where I think the CMS industry needs to broaden its scope:

• Loosen the ties to the website — As a product, TYPO3 has been fairly good at avoiding distractions, but we’re still tied closely to the website. The page tree — and pages as content containers — is a great way to organize many types of content, and it has clear performance benefits. I believe the page tree will survive. If we drop the assumption that content has to be bound for a website, we can manage content beyond the web.
   
• Create a home for the AI below the UI — A prompt is a terrible user interface, so let’s retain and improve the UI of the content management system. AI should be involved at the click of a button or automate tasks in the background. It should help us turn a PDF into structured content, translate content, classify it, suggest relationships, or replicate a page structure for all the courses at a university.
   
• Include files in the definition of content — Photos and documents have received a stepmotherly treatment. They are also content, so they should become first-class citizens of the content management system. This does not mean that every CMS should replace a specialized DAM or document management system. But it does mean bringing capabilities traditionally associated with those systems closer to content management and treating files as information with their own structure, context, and lifecycle. If content management is about managing information rather than building websites, the distinction between page content and files becomes increasingly artificial.

The CMS industry may have spent the last 30 years thinking it was providing website tools, but it’s not. A website is just one way of presenting content. The real job of a CMS is to help people understand, structure, govern, reuse, and distribute information.

AIs are changing how we create content, and they may make small websites so easy to build that there is little need for traditional website builders. But that just makes content management more important.

The CMS of the future should be a place for you and me to understand and control content, but also a repository of structured, well-governed, reusable information. This will give humans and AIs a shared understanding of the content and help make sure that it is both trustworthy and useful wherever it is needed.

Websites were fun, but never the whole story. Content management is needed now, more than ever.