Competent creative just became free. That is the most important thing to happen to brand-building in a generation, and most founders are reading it exactly backwards.
The instinct is to celebrate. Design, copy, a hundred ad variants before lunch: the cost of producing a passable version of any of it has fallen through the floor. If creative was the bottleneck, the bottleneck is gone.
But “passable” is carrying a lot of weight in that sentence, because everyone else has the same tools, pointed at the same reference material, prompted in much the same way. When the cost of competent output goes to zero, competent output stops being an advantage. It becomes the water everyone is swimming in.
The question worth your attention is no longer whether you can make the thing. It is whether anyone will remember it was you.
Generative models are averaging engines by construction. They predict the most probable next pixel, word or layout given everything they were trained on. That is exactly what you want when you need fluent, on-brief, inoffensive work. It is a quiet catastrophe when you want to be recognised.
This is measurable now, not theoretical. A 2025 study by Chaoran Liu, Tong Wang and S. Alex Yang used Italy’s 2023 ChatGPT ban as a natural experiment, tracking what happened to restaurants’ marketing when the tool vanished overnight. With access gone, their content became roughly 15% more lexically diverse and 12% more syntactically varied. Take the machine away and businesses start sounding like themselves again. Give it back and they converge.
Now scale that up. McKinsey’s 2025 survey found 71% of organisations regularly using generative AI in at least one function, with marketing and sales among the fastest to adopt it. A whole category is drawing from the same statistical centre, quietly sanding off the odd and the specific and the strange, which are the very things that make a brand stick. The output is fine. It is all fine. That is the problem. Fine is the new invisible.
There is a deeper irony underneath it. Distinctiveness is inherently a little inefficient. The memorable asset is rarely the optimal one, and the detail that makes a brand stick is usually the first thing a rational planner would cut. AI is the most powerful efficiency engine ever pointed at a marketing department. Ask it to make you memorable and it will earnestly optimise the memorability away.
Here is where founders reach for the wrong tool. Faced with a sea of sameness, the reflex is to differentiate, to hunt for a defensible product claim, a wedge, a reason you are better. That is useful for strategy and nearly useless for being remembered.
The Ehrenberg-Bass Institute has spent two decades pulling those two ideas apart, and the distinction is the whole game. Differentiation is about being meaningfully different, a rational reason to choose you. Distinctiveness is about being recognisable, the sensory shortcuts that let a buyer identify you at a glance, before a single word of your positioning has landed.
Jenni Romaniuk’s work on distinctive brand assets catalogues them: colours, logos, characters, fonts, sounds, phrases. The yellow of a taxi. The tick. The meerkat. None of these is an argument. Each is a memory hook, an asset that builds what Ehrenberg-Bass calls mental availability, the likelihood you spring to mind in a buying moment.
And here is the uncomfortable part for anyone hoping to simply out-argue the competition. Most category differences are marginal, and quickly copied. Distinctive assets, built consistently over years, are neither. They are the closest thing a brand has to a proprietary spot in someone’s head.
The science of how those assets work should make every founder sit up, because it points straight at what generative tools flatten.
Orlando Wood, chief innovation officer at System1, analysed more than 300 campaigns in the IPA’s effectiveness databank going back to 1992 and found that campaigns carrying a “fluent device,” a recurring character or scenario a brand owns over time, were far more likely to deliver market-share and profit growth than those without. His companion work on the creative itself is blunter. Reviewing 100 recent UK and US ads, he found that the more “left-brain” features an ad carried, meaning the flat, the abstract, the literal, the less likely it was to earn a strong effectiveness rating.
Read the list of what actually works: character, scene, wit, the particular. Then read what an averaging model reaches for by default: the smooth, the literal, the generically pleasant. The features that build long-term memory are precisely the features that regression to the mean sands off. AI is not neutral here. It pulls against distinctiveness by design.
When making things becomes free, all the value moves to knowing which things are worth making, and having the nerve to protect them while they still feel boring to you.
So distinctiveness is the asset. Taste is how you build it and how you defend it.
Taste is not a vibe. It is judgement: the ability to look at forty competent AI variants and see that thirty-nine of them are the category average wearing your logo, and one of them is actually you. It is the discipline to reject work that is genuinely good, because good is now abundant and ownable is not. When anyone can generate the plausible, the scarce skill is the editorial nerve to choose the specific and kill the generic.
Rory Sutherland has argued for years that brands need what he calls the right amount of weird, the deliberate idiosyncrasy of a Veuve Clicquot yellow or some strange, memorable detail that a rational optimiser would file down as inefficient. AI is the most powerful rational optimiser ever built. Left to its own devices it will remove your weird as a matter of course, because your weird is, statistically, an outlier. Someone with taste has to keep putting it back.
This is why the human does not get cheaper as the tools get better. The tools raise the floor for everyone at once, which means the floor stops being worth anything. The value is all in the ceiling now, in the judgement calls no averaging engine can make for you: what to own, what to repeat until you are sick of it, what to walk away from.
Treat your distinctive assets as infrastructure, not decoration. Pick a small, defensible set, perhaps a colour, a shape, a phrase, a character, one recognisable way of writing, and commit to them well past the point of boredom. Consistency is what turns an asset into a memory. AI makes consistency trivially cheap to produce, and therefore trivially cheap to fritter away on drift.
Use the machine for volume and speed. Never hand it the final call on what is yours. Point it at the assets you have already established and let it extend them, but do not let it invent your identity out of the category average, because the category average is exactly what it will give you.
And staff for taste. In an age of infinite competent output, the rarest and most valuable person in the building is the one who can tell the difference between work that is merely good and work that is unmistakably you, and who will fight to keep the second while everyone else drowns, comfortably, in the first.
Competent creative is free now. That was supposed to level the field. Instead it lifted the whole game up one floor, to the single thing a machine cannot generate on your behalf: the judgement to be recognisably, ownably yourself. Taste used to be an edge. Now it is the only one that cannot be copied at zero marginal cost.
The best work starts with a conversation, and a decision worth making.
Get in touch