Whilst generative AI might mean marketing teams are never lacking in content, volume and quality are not the same thing. Toby Coulthard, CPO of Jacquard, addresses why this matters and what we can do about it.

For better or worse, the idea of having a content shortage for marketing teams is now something of the past. 

To understand how we got here, it helps to rewind slightly. The rise of AI decisioning tools - which allow marketing teams to isolate the right channel, to reach the right person, at the right time - made a one-size-fits-all approach to content creation obsolete. 

With hundreds of channels requiring hundreds of bespoke content variants, copywriters and marketing teams were suddenly faced with the reality of having to manually create content for each channel. Hyper-segmentation demanded content at an unreasonable scale. 

Then came generative AI. Now, using LLMs or generic AI tools, marketing teams can easily create content for those audience segments without a drastic increase in man-hours. Essentially, AI allows marketers to scale content production like never before. The volume problem, it seemed, was solved.

But volume and quality aren't the same thing. And marketing teams have just started finding out why.

The gap between generation and optimisation 

Generation and optimisation are not the same problem. Generation is fundamentally about volume and coverage. It's concerned with producing the right number of variants across the right channels. 

Optimisation, however, is something different: it requires knowing which of these variants actually impacted people, influenced their decisions, and understanding why that was - then feeding this knowledge back into the next round of creation. Generation is an output problem; optimisation is a learning problem. The tools constructed to tackle both of these issues are almost entirely separate, and in most marketing teams, they can often operate as if they’re in different departments. 

The volume instinct

The sheer volume of content that marketing teams can produce seems to have clouded many marketers' better judgment. You'll often hear marketing teams say that prioritising the optimisation of their content can be 'done later', or that tests will be necessary when there's 'more data'. The logic is something like this: why focus on refining when you can just produce more? This instinct is understandable - but it mistakes volume for progress. 

Putting off optimisation now only makes it more difficult in the future. As segmentation becomes more granular and content demands multiply, the gap between what is generated and what is measured only continues to widen.

Why the harder problem exists

So there’s a gap: more content than ever, and less understanding than ever around what it’s actually doing. The gap exists because generation tools were built to produce, rather than to predict or learn from, their output. There's no feedback loop in most LLM-based workflows - or at least not in any real sense - content goes out, performance data sits in a separate system, and nobody connects the two. The sheer scale of the output, too, means that standard optimisation metrics are often outdated or insufficient - they’re not built to handle the scale of the output. 

This is where the limits of volume become clear. Generic tools just aren't built to offer personalisation in any meaningful way. As a result, every undetected weakness in the output compounds on itself. Volume was never the whole answer, and without a feedback loop, you can't learn what good content actually looks like at scale. As a result, teams are generating more content than ever, but accumulating very little useful knowledge about what's actually working.

What this looks like in practice - and why it matters

As we speak, we're seeing many marketing teams with sophisticated data and decisioning capabilities still sending largely undifferentiated messages. Teams may speak of differentiation, but often the differentiation in question is purely cosmetic: a word or two is changed here and there, or a certain aspect of the consumer’s identity is gestured to superficially. 

But the performance data to create real differentiation does exist. Teams have access to this data in dashboards, CRM reports, and campaign analytics. The ability to generate content exists, too. What's missing is the connective tissue between them - systems that can take what the performance data is saying and apply it to the next thing that gets written in a way that meaningfully changes the next round of content. Without that, differentiation stays cosmetic, and every new campaign essentially starts from scratch.

The window is now

The gap is still small enough that most teams aren't feeling it - but the direction of travel is clear. The increasing segmentation of brands' audiences will only become more granular and complex over time. As hyper-personalised content becomes the expectation from consumers globally, we'll only see content demands multiply exponentially - not just more content, but also for better, more personalised content. 

Brands that haven't connected generation to optimisation will soon hit a ceiling - which will look from the outside like a content problem, when in reality it’s a structural one. The policy of simply throwing more content at consumers won’t cut it when competitors offer personalisation at a much, much deeper level. 

The volume problem has been solved. The content problem hasn't - and that's precisely why now is the moment to act. It won't feel urgent yet, which is exactly what makes it dangerous. The debate on the quality of AI writing still rages - but after all this time asking ourselves whether or not it can write, we should think about shifting the question to whether or not AI knows what works. Those who aren’t asking the latter question will find themselves, all too quickly, left behind.