AI Doesn’t Need Taste – It’s Defining It

“Sure, AI can copy,” someone says. “But it can’t do taste.”

You’ve heard it before - that comforting mantra designers repeat when they feel the cold breath of automation on their neck.

Taste, we tell ourselves, is what keeps us safe. The subtlety, the intuition, the mysterious je ne sais quoi that separates the true designers from the dribblers.

But maybe that’s wishful thinking.

Maybe taste isn’t the human superpower we think it is.

Where Taste Really Comes From

Taste sounds individual. It feels like preference, personality - the design equivalent of soul.

But look closely and it’s basically exposure.

You like what you’ve seen enough times to trust. You dislike what feels unfamiliar or off-trend. You’re not born with a sense for Swiss typography, even if you are Swiss, or monochrome hero sections; you absorb it through repetition.

Taste is consensus wrapped beautifully in confidence.

It’s the collective agreement (mass hysteria?) that certain shapes, colours, and typefaces belong together - until they don’t.

And that consensus is shaped by whatever we’re most exposed to.

For the last few decades, that meant magazines, conferences, awards sites, Dribbble, Behance, and whatever Apple was doing this year.

(Sadly I don’t think Google are getting enough props for their work on Material to be on our little list here.)

But exposure has new curators now.

The Algorithm As Tastemaker

Recommendation engines already decide what we see - from Spotify to TikTok to Pinterest.

They reward engagement, not originality.

They don’t understand why we like something - only that we do.

AI training models work the same way. They’re built on the data of what’s already popular, and they reproduce it beautifully. Not because they have taste, but because they’ve inhaled it.

Every time an AI tool generates another “clean, modern landing page”, it’s feeding more of that look back into the ecosystem.

More exposure.

More consensus.

And there’s a lot of it. In April 2025, roughly 74% of all new English-language web pages contained AI-generated text or imagery. Three in four. Three in four!

That means most of what we now see online is algorithmic echo - and exposure, remember, is how taste forms.

The machine doesn’t need to understand taste. It just needs to saturate the environment with its output until its version of taste becomes the baseline and it’s output is simply exhaling what it has inhaled.

The Beige Web

You can already see it happening.

Tailwind CSS, design systems, and component libraries have made websites cleaner, faster, more consistent - and almost indistinguishable from one another.

AI just puts that process on steroids.

A prompt like “modern SaaS homepage” will give you a perfectly nice layout in seconds. Soft gradients. Rounded cards. Muted blues. Trustworthy sans-serif.

It’s tasteful, sure.

It’s also identical to everything else.

Taste, as it turns out, is a flattening force.

Once you automate “good design”, you automate the average.

This isn’t a criticism of AI or Tailwind - they’re efficient responses to a market that values reliability over risk. We used to say nobody got sacked for buying IBM, perhaps now some are thinking nobody got sacked for using Tailwind.

But they illustrate how “taste” - once a marker of discernment - can easily become a form of conformity.

A recent Adobe survey found 63% of creative professionals already worry that generative AI is leading to homogeneous content that doesn’t stand out.

So it’s not just visible. It’s measurable.

The Comfort Of The Familiar

Of course, sameness isn’t always a sin.

Jakob’s Law - the idea that users prefer your site to work like every other site they already know - is one of UX’s most enduring truths. I’m no fan of Jakob’s long running campaign to “beige” the web but there is truth in what he says - even if it is an uncomfortable one.

In practice, that means they perform tasks faster and with less friction when an interface behaves the way they already expect it to.

Consistency is usability.

It’s cognitive empathy in code.

Research backs it up: studies show that familiarity improves both speed and accuracy when users interact with digital systems (NN/g, arXiv).

Familiarity isn’t just comfort — it’s performance.So when Tailwind, design systems, and now AI make everything feel comfortably alike, that’s not just laziness - it’s service design.

You’re saving people time, friction, and mental load.

But sameness is also brand kryptonite.

When everyone uses the same components, differentiation starts to live only in tone, motion, or copy.

You’re not designing a thing, you’re designing a flavour.

As Rory Sutherland puts it, “The fatal issue is that logic always gets you to exactly the same place as your competitors.”

In other words, breaking the pattern - when done with wit and intent - can be just as valuable as following it.

Because comfort builds trust, but contrast builds memory.

AI’s growing grip on taste will make the familiar frictionless - but it’ll take a real human to make it memorable again.

Exposure Loops

Designers feed on what they see.

We scroll through the same interfaces, pin the same palettes, copy the same grids.

That’s how styles propagate - via osmosis.

Now AI models are trained on that same visual soup, and we’re using their output as reference material.

It’s a perfect feedback loop: humans teach machines what’s tasteful; machines produce more of it; humans learn from the machines.

And the loop is tightening. A 2024 Adobe survey found 83% of creative professionals already use generative AI tools in their workflow. Those same creative professionals that worry about AI driven beige (#f5f5dc).

That means the material training tomorrow’s models is already touched by today’s models.

Soon, our sense of what “looks right” will be calibrated to what AI produces.

We’ll call it evolution, but it’s really reinforcement.

And when everyone’s exposed to the same aesthetic feedback loop, individuality becomes an outlier - something the algorithm quietly pushes out of frame.

The Myth Of Human Taste

So maybe “AI can’t do taste” isn’t quite the flex we think it is.

Because taste was never mystical. It was just slow.

It took time - years of exposure, reflection, iteration - for consensus to emerge.

AI compresses that timeline into seconds.

What feels like a lack of taste is really just the absence of delay.

Humans need time to notice trends and copy them. Machines don’t.

By some estimates, up to 90% of online content could be synthetically generated by 2026.

At that point, “popular” and “machine-made” will be the same thing.

When you flood the cultural landscape with AI-generated work, the lag between “inspired by” and “derivative of” disappears.

The machine doesn’t steal your taste. It accelerates it until it collapses into itself.

Designers In The Age Of Beige

This is the part where I’m supposed to argue that designers will still matter - that human taste will somehow triumph.

And yes, maybe it will. But not because it’s sacred.

Designers will matter if they can stay weird.

If they resist the comfort of consensus.

If they use AI not to polish the average, but to break it.

The danger isn’t that AI replaces taste - it’s that it replaces curiosity.

If we stop questioning why things look the way they do, we become stylists for a machine’s idea of elegance.

That’s the quiet trap of tools like Tailwind and AI builders: they make the right thing effortless, but the interesting thing harder to justify.

Escaping The Echo

So what do you do when everything starts to look the same?

You can’t out-AI the AI. You can’t avoid exposure. And you can’t pretend the trends aren’t shaping you too.

But you can decide where you take your cues from.

Designers often forget that creativity doesn’t start in Figma. It starts in seeing differently. When the visual field narrows, the trick is to widen your inputs.

Look sideways, not up.

Study adjacent professions - architecture, typography, stage design, packaging, ceramics, theatre, service even gardening. They each have their own logic of beauty and constraint. You’ll start spotting principles that apply to digital work in unexpected ways.

And step away from screens altogether. Museums. Second-hand books. Supermarkets. Pavements. The texture of the world is full of interface ideas hiding in plain sight.

The point isn’t to reject the mainstream aesthetic, it’s to dilute its dominance.

You might look at other professions such as service design to learn what they value as desirable, or quality. The answers might surprise and shape your thoughts about to apply aesthetics.

Feed your brain something the algorithms can’t recommend.

Because originality doesn’t come from isolation - it comes from exposure to different things.

Who Defines Taste Now?

If taste is exposure, then whoever controls exposure controls taste.

That used to mean editors, curators, and award judges.

Now it means algorithms.

The feed decides.

The prompt decides.

And increasingly, the machine decides what we even get to see long before we call it “good”.

It doesn’t understand taste, but it doesn’t need to.

It sets the parameters for what we call tasteful simply by being the dominant producer.

When everyone’s drawing from the same dataset, taste becomes less about choice and more about gravity - a pull toward the mean.

Maybe That’s Fine. Maybe Not.

Every generation mourns the loss of something they once thought was theirs alone.

Photography didn’t kill art. Sampling didn’t kill music.

AI won’t kill design either.

But it will change the social mechanics of taste - who defines it, who reinforces it, who profits from it.

And that shift is already underway, quietly, beneath the glossy interfaces we call “tools”.

If we want to keep taste human, we have to remember that it’s a social construct - and we still get to decide what we expose ourselves to.

Maybe that’s the new creative act: curating your own input in a world where the machine would love to do it for you.

AI doesn’t need to have taste.

It just needs to make everything tasteful enough that we forget why difference mattered.

Call to Action

If this got you thinking about how your own sense of taste is shaped (and possibly hijacked), drop a comment below.

What’s the last thing you created purely because you liked it - not because a prompt, a trend or a stylesheet told you to?

Let’s talk about the weird, the off-beat and the human touch that AI can’t (yet) replicate.

Sources

• “74.2% of newly created English-language web pages in April 2025 contained AI-generated content.” — Ahrefs/StanVentures analysis. ahrefs.com
• “83% of creative professionals said they’re using generative AI tools in their work.” — Adobe blog, 2024. blog.adobe.com
• “63% of creative professionals worry that generative AI will lead to homogeneous content that does not stand out.” — Adobe survey, cited in Superside article. superside.com
• “Up to 90% of online content may be synthetically generated by 2026.” — Oodaloop forecast. medium.com

Author’s Note

Written by a human. Probably.