It is said that anything can now be created with AI. I believe that is true. Recently, even without engineering knowledge, I was able to create and launch an app. Banners, proposal materials, and copy are easy; even videos and entire websites can be produced instantly if you can verbalize your intent. The era where design and verbalization skills functioned as a barrier to entry is rapidly coming to an end.
Costs are down, speed is up, and even businesses with few resources can obtain a certain quality of output. In fact, many companies that do not make a living from design perceive it this way. However, from a branding perspective, there is a serious problem slowly progressing here.
The 107% Cheerful Average
A study recently featured in WIRED strongly pointed this out. According to joint research by Imperial College London, Stanford, and the Internet Archive, approximately 35% of websites newly published since 2022 were written by AI (or with AI assistance).
https://wired.jp/article/ai-slop-is-changing-the-internet-just-not-how-you-might-think/
What was surprising was that most of the research team's prior hypothesesâthat AI would make writing uniform and mechanicalâwere wrong. The writing style was not as homogenized as expected. Fake news and misinformation did not increase as much as anticipated either.
Instead, two clear results emerged. The positive emotion score of AI-generated websites was about 107% higher than that of human-written sites. The writing was uncannily "cheerful." The research team described this as a symptom of "sycophancy" and "excessive optimism" in large language models. Furthermore, the semantic similarity score was about 33% higher. It wasn't the style, but the diversity of the ideas themselves that was being lost.
Why Do the Shops That Rejected Chains All Look the Same?
Interestingly, this homogenization isn't just about text on the internet.
Around the same time, there was another study published in The Conversation. Researchers from the University at Buffalo and the University of New Orleans surveyed over 100 young professionals in North American cities about the interior photos and design elements of their "favorite indie coffee shops."
The results were brutally consistent. In about two-thirds of the favorite shops, baristas had tattoos and piercings, and there were exposed brick walls, chalkboard menus, reclaimed wood counters, local art displays, vintage furniture, and houseplants. The symbols of being "indie, authentic, and cool" were as common as if they had been stamped out by a press.
The research team further showed six photos and asked participants to guess whether the cafe was in Cincinnati, St. Louis, or Toronto. No one got them all right. In a test to distinguish between just two cities, Chicago and San Francisco, only 6% could correctly identify both. Individually owned shops that were supposed to reject the uniformity of chains were visually indistinguishable, even when 3,000 km apart.
In other words, the shops that started as independents because they hated the uniformity of Starbucks ended up becoming "just another chain."
Two Studies Saying the Same Thing
The two studiesâone on the AI-generated web and the other on North American indie coffee shopsâseem to deal with different phenomena, but they are essentially saying the same thing.
When each individual continues to choose what is considered "rationally good," the result for society as a whole is that everyone converges toward a similar "pleasant average."
Cafe owners wanted to make their shops attractive. They didn't choose similar interiors out of malice. They referenced "well-regarded looks" on Instagram and Pinterest, and to reduce the risk for bank loans, they chose safe designs that would appeal to a wide range of customers. As a result, the sum of each person's optimization appeared as the homogenization of the entire city.
AI slop follows the same structure. The person in charge isn't trying to cut corners. In fact, now that they can use AI, they are working harder by expanding beyond their usual domain. By adopting the "positive, smooth, and likely-to-engage expressions" suggested by AI one by one, corporate social media, proposals, and websites become something uncannily cheerful and uncannily similar.
This is exactly the problem facing branding today. Rational optimization by individuals creates homogenization as a group. And AI increases the speed and scale of that homogenization by orders of magnitude.
Slop is a Slow Death
These outputs mass-produced by AI are called "slop," referring to low-quality generative content in general. However, this isn't an obvious defect. It's not a typo, a grammatical collapse, or a blatant lie. On the surface, it's clean, bright, and pleasant. There's no friction while reading it. Yet, after you finish reading, nothing remains. The sense of who the words were for has vanished.
The real problem with slop appears when it accumulates. It's similar to the kind of damage caused by dying from eating nothing but sweet, cute snacks. Every bite is delicious. There is a sense of satisfaction. But the person is unlikely to notice that the accumulation is eroding the very structure of their body over many years. You can spit out spicy food or something rotten on the spot. You can't spit out snacks.
This is exactly what is happening in corporate branding. Daily social media posts, banners, proposals, internal documents, store interiors, and product packaging. The accumulation of outputs that aren't bad individuallyâand are actually quite pleasantâis quietly dissolving the outline of the brand. The person in charge thinks they are doing something good, but they are bringing the brand closer to death every day.
While rejoicing that "AI made it cheap and fast" or "It's safe because I referenced Pinterest or Mobbin," the distinctiveness and uniqueness that the company has built up over the yearsâthe most essential assets of the brandâare depreciating a little every day. It's not a single major accident, but a slow death where small daily shortcuts accumulate like geological layers.
A Brand is the Accumulation of "What You Dared Not to Do"
I strongly resonated with a post recently written by Takanori Kataishi (CEO of Yutori Co., Ltd.) on X.
I believe this is one of the most important definitions when talking about branding in the AI era.
What AI is good at is precisely providing the optimal solution for "what you should do if you think rationally." SEO-strong headlines, syntax likely to get engagement, appeals that seem to hit the target, color schemes likely to win A/B tests. AI has speed and precision that overwhelmingly exceeds humans when it comes to maximizing rationality.
However, the outline of a brand is created on the opposite side. Even if you know it's rationally better to do something, you dare not do it. You don't go for short-term engagement. You don't match the trendy tone. You don't over-explain. You don't choose words that appeal to everyone. The accumulation of these "dare not to do" decisions creates a brand that can be distinguished from others.
Conversely, the more you surrender to the rationality of AI, the more the brand converges to the average recommended by the algorithm. That average that is 107% cheerful and 33% similar. That average of exposed brick walls, tattooed baristas, and chalkboard menus.
In other words, the essence of branding in the AI era is not "what to make the AI do," but "where you can refuse what the AI rationally recommends."
Three Abilities That Constitute the Power to Refuse
The abilities needed to refuse AI recommendations and maintain your company's outline can be organized into three categories.
First is Taste.
This is judgment, an accumulation of knowledge and experience. You could call it the muscle for deciding what not to choose, even more than what to choose. AI will provide infinite options, but to select the "correct answer for your company" and discard the 80% that is rationally recommended, you need a reference axis for judgment.
This can only be trained through curiosity, constant input, and rumination. Even more important is the ability to apply a different context to the main context and read the meaning from the gap. When thinking about visuals for a beverage brand, no difference will be born if you only look at the beverage category. Architecture, contemporary art, street culture, photo books of local festivals. There are lines between the lines that only emerge when you line up seemingly unrelated things. A person who can capture those lines can make taste judgments that exceed the AI average.
Taste is often lumped under the vague word "sense," but its reality is steady accumulation and rumination. It is an asset that needs to be nurtured over time as an organization.
Second is Will and Courage.
Will is, in short, deciding "what to put with what." The combination of copy and visuals, the shifting of categories and expressions, the judgment of when to release and when not to. AI presents countless combinations, but deciding "this is the one" is a job that can only be placed on the human side.
And will is always accompanied by courage. It always takes courage to put something into the world that doesn't exist yet. It's easier to get internal consensus for the "positive, smooth average" that AI suggests first. It's easier to find a "well-regarded look" on Pinterest. Choices that deviate from that always come with the anxiety of "is this really okay?"
Whether a brand can have distinctiveness and unique appeal ultimately depends on whether the organization can "endure this anxiety." Will and courage are matters of organizational character, not technology, and the AI utilization of companies that haven't trained this will surely flow in the direction of slop.
Third is Humor.
Humor is a device for sharing values that lie outside of language by broadcasting that "this is interesting," thereby meeting and connecting with someone. Someone who shares the same points of interest sees the world the same way you do. Conversely, someone who doesn't get it is living on a different layer altogether.
What's important is that the sender has the resolve to say, "It's fine if people who don't get it, don't get it." The moment you add explanations so everyone understands, humor dissolves into the average. Only through active filteringâdaring not to reach everyoneâcan a strong circuit of empathy be born with the intended audience.
This is also the area where AI is weakest. LLMs are trained to behave so that they are understood by everyone, and the judgment that "it's fine if people who don't get it, don't get it" will never appear in an AI's default output. The reason the advertising for Liquid Death, Oatly, and Duolingo works is not because it's funny, but because through that funniness, they are selecting and bundling "their people."
Using humor as a weapon is not about being light. It is an extremely strategic act of accepting that "it's fine if those who don't get it don't get it" to connect deeply only with those you truly want to reach.
Good Medicine Tastes Bitter
Taste, will and courage, and humor. These three will not become commoditized even as AI versions improve. This is because they do not reside in tools, but only in the bodies and experiences of the humans within an organization.
Conversely, an organization that hasn't nurtured these will dilute its own brand the more it introduces AI. It's the same situation as eating snacks every day and convincing yourself it's a healthy diet. What happened to North American indie coffee shops will happen in every industry. And with AI, the speed and scale will be orders of magnitude greater. Before you know it, your company's creative will be indistinguishable from competitors, unrelated startups, or shops in other cities 3,000 km away.
The responsibility of those in charge of branding lies here as well.
The agency model that moves only after receiving a request from a client is no longer fast enough. By the time a client feels a sense of "vague unease," the internal AI has already produced a cheerful, "plausible-looking" proposal. Even if you intervene with a standard approach from there, the ground for discussion has already been pulled toward the average.
And it's not that the clients themselves are bad. Rather, they are just rationally and seriously continuing to optimize. The problem lies in the structure where those "rational daily choices" directly become the consumption of snacks. It's exactly the same as the North American cafe owners who chose similar interiors without any malice.
The role of external design firms and branding partners changes here. In a world where "pleasant things" can be obtained in any quantity through AI, there is almost no value left in being a "person who makes pretty things." What remains is whether you can consciously take the position of prescribing "bitter medicine."
It's hard to swallow in the short term. Internal consensus is hard to get. Initial engagement might be lower than the average. But five or ten years from now, it is definitely the choices on this side that will be creating the outline of the brand. You move to place the concrete details first so you can say that with certainty. The prototype itselfâsaying "rationally it's this, but we're daring to go with that"âdefines the ground for discussion. It's a structural matter where the one who places the concrete details first gets to draw the brand's outline.
AI can already act as the entity that hands out snacks. That's why design firms from now on have no choice but to move to the side that hands out bitter medicine, along with the reason why it must be taken now.
What remains for branding in the AI era is not technology. It is whether an organization can continue to choose its own "bitterness" without being swept away by daily rationality. And for those supporting them from the outside, it is whether they have the resolve and grit to keep providing that bitter medicine.
