journal · 20 aug 2026

why AI brands all look the same

Open ten AI product websites in a row. Same gradient. Same violet-to-cyan glow. Same sans-serif set in the same weight. Same abstract 3D shape floating over the same dark hero.

There’s a reason, and it isn’t laziness. It’s speed. When a category moves this fast, copying the visual conventions of the category is the safest available decision. Nobody gets fired for looking like the last company that raised.

The problem shows up later, at the point where a buyer has to remember which one you were.

sameness is now the default output

Two forces push every AI brand toward the same picture.

The first is the reference pool. Founders brief their designers with competitor screenshots. Designers pull moodboards from the same three sites. The category feeds on itself and tightens with every iteration.

The second is generative tooling. Ask any image model for “modern AI startup brand” and it returns the statistical average of everything it saw. Averages are, by definition, the middle of the pile. Use that output as your identity and you’ve automated your way into the crowd.

Both forces produce work that looks competent. Competent and interchangeable.

what happens when a company opts out

Granola, an AI notepad, rebranded against nearly every convention of its category. A deliberately imperfect logo. A typeface built from the actual handwriting of co-founder Sam Stephenson. A warm earthy green chosen to sit apart from the bright RGB hues everywhere else in tech. Nothing about it reads as an AI product, and that’s the point.

The results are on the record. Launch day produced the biggest download day in the company’s history. A month later Ramp data ranked Granola the second-fastest-growing software brand in the world. Three months later it raised $125m at a $1.5bn valuation, with investors citing the brand as a differentiator.

One case doesn’t prove a rule. It does show that “we can’t afford brand risk right now” has a cost on the other side of the ledger, and that cost stays invisible until a competitor collects it.

the same problem outside AI products

Fintech has its own version of category sameness: the same blues, the same gradient card floating at an angle, the same reassuring roundness. The pressure is identical, and so is the outcome.

We rebranded Paysend together with their own design team, a fintech moving money across 170+ countries. The business had scaled well past the brand it was still using, which is the normal sequence: product and growth first, communications catching up later.

Interviews with the C-level and category research produced one insight: Paysend should speak as the infrastructure behind global payments, not as another consumer money app. That single line decided the rest. The visual metaphor became an infrastructure grid. The logo reduced to three strokes, one mark that works whether the audience is a bank, a partner or someone sending money home.

Then the part fintech convention would never allow. Instead of one corporate blue, the palette runs 15 colours and splits by product across five audiences, so consumer, enterprise, ambassador, investor and partner each get their own register of the same brand. A payments company that looks like a payments company picks the blue. This one didn’t need to, because the insight said infrastructure, and infrastructure carries many things at once.

where distinctiveness actually comes from

Every element in both cases traces back to something true about the company. A co-founder’s handwriting. A product built around presence in meetings. An insight, pulled out of a company’s own executives, about what business it’s actually in.

You can’t generate that. A model has no access to your founder’s handwriting, your origin story, the thing your team argues about, the reason your first ten customers stayed. Those are inputs only you have.

This is the dividing line in AI-era branding. The identity comes from proprietary material about a specific company. AI carries the production: the variations, the resizes, the localisations, the thousand assets that identity needs once it exists.

We run our studio exactly this way. AI is deep in our pipeline, and it’s nowhere near the part where we decide what a brand is.

how to check your own brand

Four questions, answerable in ten minutes.

  1. Cover the logo on your homepage. Could it be your closest competitor’s site? If yes, you have a memorability problem, whatever your traffic numbers say.
  2. Where does your visual system come from? If the honest answer is “category conventions,” the system is borrowed.
  3. What in your brand could only belong to your company? Name one thing. If the list is empty, that’s the brief.
  4. What are you refusing? Distinctive brands are defined as much by the thing everyone else does that you won’t. Granola refused sleekness.

the trade

Standing out means looking unfamiliar for a while. Internal stakeholders will ask why it doesn’t look like the companies they admire, and that question is the whole test.

The alternative is a brand that clears every internal review, matches the category, and gets forgotten between one demo call and the next.

Both options cost money. Only one of them compounds.

Granola figures via Design Week and Creative Boom. Paysend rebrand: our case.

UNKNW is an AI-native creative studio in London and Barcelona. We work with growing companies.

published: 20 august 2026