Recognizable, or invisible. Brand language in the age of AI slop

A banner, a landing page, a product visual, a LinkedIn post. Today you can have all of them in an afternoon, and all of them will look fine. Clean type, soft gradients, a confident headline. The trouble is that your competitor has the same afternoon, the same tools and very often the same prompt.

Slop got its own word

In December 2025, Merriam-Webster picked “slop” as its word of the year and defined it as low-quality digital content produced in quantity by AI. Australia's Macquarie Dictionary went straight for “AI slop”.

It is not just a feeling. When Kapwing set up a brand-new YouTube account and watched the first 500 Shorts it was served, 21% of them were AI-generated slop.

Most of what companies publish is not slop in that sense. But it lives in the same feeds, next to it, and people's eyes are learning to skip anything that looks generated.

Average by design

Generative models are very good at producing the most probable answer. That is what makes them useful, and it is also why their output drifts toward the middle.

Researchers are starting to measure it. In a study in Science Advances, Anil Doshi and Oliver Hauser found that writers with AI ideas produced stories rated as more creative, yet the stories became more similar to each other. Individually better, collectively narrower.

The Artificial Hivemind paper, presented at NeurIPS 2025, tested more than 70 language models on open-ended questions. Not only does a single model keep repeating itself, different models from different companies give strikingly similar answers.

Translate that into brand work. If a bank, an e-shop and a SaaS startup all ask for “a modern, trustworthy, friendly visual”, they will get variations of the same thing. Design had a sameness problem long before AI. People called it blanding: the same sans-serif logos, the same moodboards. AI just made the average cheaper and faster than ever.

Recognition is a memory, not a style

Jenni Romaniuk from the Ehrenberg-Bass Institute has spent years researching distinctive brand assets: colors, shapes, characters, sounds, ways of speaking. She measures them on two axes: how many category buyers link the asset to the brand (fame), and whether they link it only to that brand (uniqueness).

Two of her points matter here. Recognition is built through long, consistent use, and frequent “refreshes” tend to push a brand back rather than forward. And marketers routinely overestimate how strong and unique their own assets are.

A generic asset has zero uniqueness by definition. If your visual language is what a model would produce for anyone in your category, you are not building memory. You are paying for someone else's.

NIQ saw something similar from the other side. In their research on AI-generated ads with more than 2,000 participants, even the ads viewers rated as high quality triggered weaker memory activation than traditional ones. It is vendor research, so take it with a grain of salt. The direction is still worth noticing.

The Coca-Cola lesson

Coca-Cola's AI-generated Christmas ads drew heavy criticism online in 2024 and again in 2025. Creepy faces, odd motion, “soulless”. Yet testing firm System1 gave the 2024 version a top score.

How can both be true? The ad was carried by assets the brand has kept consistent for decades. The red trucks, the lights, the “Holidays Are Coming” melody. AI produced the frames, but recognition came from a memory built over thirty years.

A brand without that reserve would have produced one more pleasant clip that nobody connects to anyone. AI amplifies what is already there. If there is nothing distinctive, it amplifies the average.

What design maturity actually means

Design maturity is not the size of a design team or the number of tools. It is a company's ability to decide how it looks, sounds and behaves, and then keep to it every day, across people, agencies and now machines too.

In practice that tends to mean:

1. A brand language described clearly enough to hand over. Not just a logo manual, but principles, tone of voice with real examples, and a list of things the brand never does. If it cannot be explained to a new colleague, it cannot be explained to a model.
2. A design system as the shared source of truth. Tokens, components and rules that make the right choice the easiest one, whether a person or an AI tool is producing the output.
3. Someone who owns it. Consistency does not happen by itself. Someone has to have the authority to say “this isn't us”, even when it looks good.
4. Restraint with refreshes. Before changing a color or a font, ask what people actually recognize you by. Change the generic parts, protect the distinctive ones.
5. AI as a tool, not an author. Let it speed up production, variants and testing. The decision about what the brand is stays with people.

Speed is a commodity

Anyone can generate. Not everyone knows what to keep, what to throw away, and why this one fits the brand while the other nine do not. That ability does not come with a subscription. It grows over years of consistent decisions.

In a landscape where the average costs almost nothing, being recognizable becomes one of the few real advantages left. And it is built the same way it always has been: slowly, consistently, and deliberately. Less, but better.