The (content) king is back

Why the flood changed the math  

Generative AI removed the volume constraint overnight. The cost of producing a paragraph, a product description, a campaign variant, dropped close to zero. The intuitive conclusion is that content is now cheap, abundant, and therefore worth less.  

That conclusion is exactly backwards.  

When anyone can generate infinite output, production stops being the valuable part. What becomes scarce is everything production cannot do on its own: accuracy, structure, reuse, the ability to take one piece of content and deploy it across a website, an email, a support flow, and an AI assistant without it falling apart. Generating content is free now. Generating content you can actually trust is not. That gap is where all the value went.   

When production becomes free, value moves to whatever production cannot create on its own: structure, trust, and the ability to be used everywhere.   

The crown was never on content. It was on structure.  

This is the part the industry keeps missing, and it is worth being blunt about. "Content is king" was never really about content in the loose sense, the words, the images, the posts. It was about structured content: content that knows what it is. A price tagged as a price. A product name in a field, not buried in a sentence. Guidance modeled so a system can retrieve it, recombine it, and serve it in a context the author never anticipated.  

For years, structuring content was the corner everyone cut to hit the deadline. It felt like overhead. It felt optional. That was a bet that the foundation would never be tested, and AI just called the bet.  

And the foundation is not in the shape we pretended it was. In Grant Thornton's 2026 AI Impact Survey, 55% of technology leaders said fewer than half of their core applications were even ready for AI, with data readiness named one of the leading causes of AI underperformance. We spent a decade building the experience layer on top of a base most organizations now admit was never prepared to carry it.  

Agents do not read. They retrieve.  

Here is the reframe that should change how you plan. An AI agent does not browse your site the way a person does, skimming, inferring, forgiving the gaps. It retrieves. It needs the right fact, in the right form, and it needs to trust that the fact is current and correct.  

Hand an agent a kingdom of unstructured blobs and you get precisely what you should expect: contradictions, hallucinations, and a customer interaction you never approved and cannot trace. Hand it structured, governed content and it becomes the thing everyone is promising on stage.  

So the experience we spent a decade perfecting turns out to be the output, not the engine. The engine is the structured content underneath it. We fell in love with the output and starved the engine, and AI did not break the engine. It just turned the lights on so we could finally see the state of it.  

Stop optimizing the throne room  

If you are planning for an agentic future, the instinct is to start at the agent. That instinct is part of the same mistake. Start one layer down.  

Audit whether your content is modeled or merely stored. Most organizations discover, uncomfortably, that they have been storing. Ask whether a piece of content can survive being reused in a context it was never written for, because agents will put it in exactly those contexts. Decide who owns content quality and governance now, before an agent ships something on your behalf that you would never have signed off on.  

None of this is glamorous. It is the unfashionable, foundational work that a decade of experience-layer excitement gave everyone permission to defer. But the teams that win the next few years are not the ones with the most ambitious agent strategy. They are the ones whose foundation can actually hold one up.