Marketing’s next AI problem isn’t creativity, it’s taste

Originally published in creativebrief

As marketers, we shouldn’t be worrying about AI slop.

Instead, the real ‘threat’ is that AI is starting to produce convincing work. Convincing, but not new.

A generative model can only recombine what already exists, so what comes back is confident and coherent, but built entirely out of the past. Give it a brief, a few references and the right prompts, and it can return something polished and recognisably ‘on brand’ in seconds. It’s followed the brand guidelines and is ready to be put in front of the client.

Impressive, sure. But it is also creating a problem for agencies. Work may look finished, but the thinking behind the idea has never really begun.

The danger is not that those alternatives will necessarily be bad. Anyone can reject a bad idea.

But when that ‘harder’ brief comes in, it can be tempting for a busy team to turn to the machine without asking questions… What does it say? Why should people care? Are there different directions we can take? There’s a real risk of just accepting what is in front of you rather than questioning, exploring, experimenting or pushing back.

Questioning what’s in front of you is at the heart of creative judgement.

Creative judgement develops through doing the work: forming an opinion, trying an idea, getting it wrong, experimenting, pushing it further and gradually learning the difference between something that works and something that merely looks like it should.

It can be slow and uncomfortable, but it’s also how taste is built. And the risk of losing it isn’t one mediocre campaign. It’s that everyone ends up working from the same models, the same references and the same defaults, and the whole category sails towards the ordinary.

Taste is a skill

Taste is often treated as something instinctive. Someone ‘has it’ or they don’t. But creative judgement is developed through exposure, observation and experience.

You study great work from senior creatives and understand why it works. You defend your own ideas, receive feedback and learn how creative decisions perform in the real world. Over time, those experiences help you recognise the difference between something familiar and something genuinely strategic, original, and compelling.

Taste isn’t only how you judge the work in front of you, either. It’s what allows you to arrive with a point of view in the first place – a read on the audience, the culture or the category that’s worth defending. That’s the part that can’t be prompted, because it doesn’t exist in the training data until someone has it.

None of that happens in a single prompt.

That doesn’t mean juniors should be kept away from AI. Quite the opposite. The question is whether its use helps them develop judgement or allows them to avoid exercising it.

An experienced creative can look at multiple routes that an AI has taken and recognise that, for example, 80% are variations of the same idea. They may spot a small detail in 1% of the ideas that is worth developing, while seeing that the remaining 19% of AI outputs don’t even answer the brief. But they can do that because they knew what they were looking for before they prompted. Without a point of view going in, you aren’t directing the work – you’re simply picking the least familiar option from a list of familiar ones.

More options do not mean you have more choice.

It can also be the space in which unexpected connections emerge and an idea becomes something you could not have arrived at by taking the most obvious route.

That’s the training ground worth protecting: an environment where great work is studied, taste is passed down, conventions are questioned, and confidence is built. You don’t get that confidence from a chatbot.

Lose this, and agencies will have more ideas than ever, but fewer effective, original ones.

This is a taste problem, not an ideas problem

The industry does not need AI to produce more ideas. The harder part is arriving with something worth pursuing, and knowing why it deserves another hour, another round, or a place in front of an audience. And that’s where taste becomes commercially important.

When AI can quickly generate work that follows established conventions, the ability to distinguish between familiar and right, or polished and interesting, becomes incredibly important. Because brands grow by being easy to notice and easy to recall when someone is finally in the market to buy – and that only happens when the work is ownable. Work that follows the conventions of its category doesn’t just fail to impress. It fails to register, or it gets credited to the brand that got there first.

This is about more than just the development of creative teams. If agencies are using similar tools to produce work from similar references and briefs, all ideas are going to look the same. It may answer the brief, it may hit the brand guidelines, and it may have that polish. But it isn’t going to distinguish you from everyone else answering the brief.

That sameness might not always be obvious. Two campaigns can look different on the surface, but the same underlying connections are there, the same tone is there, and the same ‘safeness’ is there. If you have taste, then a creative team will notice when an idea is merely well executed and when it has something distinctive to say.

None of this is a case for originality for its own sake. It’s originality in service of ownability – work that could only belong to that brand.

Again, this isn’t a call to remove AI from the whole process. The tech can help find that distinction. It can show the variations of an idea, challenge assumptions, or help you realise that an original idea is actually possible. The value comes from what the team, what the creative, what the marketer, does with that information.

Settle with the ‘that will do’ answer? Or push it further?

So what should creative leaders do?

This isn’t a question of whether you should or shouldn’t use AI. Instead, creative leaders need to be deliberate about the decisions people are learning to make, not simply where they are allowed to use AI.

A blanket rule of ‘think first, prompt later’ might seem appealing, and could be useful on some briefs. But it isn’t going to teach anyone good judgement or develop taste. Not by itself, anyway.

A better approach would be to consider how work is reviewed. When a team presents an AI-assisted route, ask them what they were trying to solve before they prompted. What alternatives were rejected? Why were they rejected? Why this idea? What was the thinking? If the answer is surface-level in that it looks polished, then there is more underlying thinking to be done.

Junior creatives also need to be part of those discussions, rather than seeing only the final decision. Hearing a senior creative explain why an apparently strong route feels predictable, or why a rougher idea is worth another day’s work, is how taste gets passed on. Critique matters as much as the output.

AI is a remarkable instrument and a poor author. Used to widen what we explore, it makes us faster and braver. Used to decide what we think, it makes us all sound the same.

Agencies will get the most from it when they protect the experiences that help people form an opinion on work, to critique it, to develop taste, allow them to fulfil their potential, and deliver more compelling, original, effective work.