Everyone is asking what AI will save them. Fewer people are asking what it does to the thing they actually sell.
The comfortable story: AI makes marketing cheaper, so you do more of it, so you win. The less comfortable version: AI makes competent marketing cheap for everyone at the same moment. When something becomes free for all, it stops being an advantage for any.
Competent used to cost money. A decent headline, a clean landing page, a tidy deck: all of it needed someone with a few years of practice and an invoice to match. Now it needs a prompt. If your pitch is "solid, professional work, delivered fast", you're selling the one thing the machine does best, at a price it undercuts by default.
That's an argument, not a finding. But there's some evidence under it, and it's worth being precise about how much.
The machines agree with each other
n 2026, Emily Wenger (Duke) and Yoed Kenett (Technion) published "Large language models are homogeneously creative" in PNAS Nexus. They put 22 language models and 102 people through three standard creativity tests.
On originality per answer, the machines held their own: roughly level with the humans. The gap was elsewhere. Ask the models the same question and the answers cluster. Ask people and they scatter. On one test, the authors' variety score (higher means more varied) was 0.459 for the models and 0.699 for the humans. They conclude it's a property of language models in general, not of one model.
Two caveats. These were lab tasks, like inventing unusual uses for an object. Nobody tested ad copy or positioning. So the study doesn't show your AI-written campaign looks like your competitor's. It shows the tools you both use are individually decent and collectively samey. The step from there to your marketing is ours, and it's a short one, but it is a step.
Polished is the entry fee
DoubleVerify's 2026 EMEA media quality report (August) is a survey, and the coverage doesn't give the sample size, so treat it as a mood, not a measurement.
Across EMEA, 42% of consumers said they'd feel negatively about a brand using AI for ads that look low-quality, uncanny or strange; in the UK, 49%. When the AI ads looked polished and professional, 29% of UK respondents reacted negatively.
That doesn't say people love AI ads. It says sloppy AI work gets punished and tidy AI work mostly doesn't. Tidy is the floor. Nobody gets credit for standing on it.
What stays expensive
If competent is free, the premium moves to what the model doesn't hand you by default. A position, to begin with: a claim someone could disagree with. Quality, innovation and passion are claimed by everyone. Then a choice you can defend: this audience, not that one, this channel, not five. And a voice with some risk in it. If a sentence could sit unchanged on a competitor's site, it probably will.
The test
Take your last three pieces of content and swap in a competitor's name. If nothing breaks, you didn't have a position to begin with. AI didn't take it; it just made the gap visible, faster.
Use the model for volume, drafts and variations. Spend the hours it saves on the decision upstream: what you stand for, who you're not for, what you won't say. When average costs nothing, the only thing worth paying for is the reason to pick you.
