Why human-made patterns are the foundation of good AI design

When people see an AI tool produce a tidy, well-proportioned web page, it can look as if the machine has invented good taste. It hasn't. What it has done is apply, very quickly, a body of knowledge that people built up slowly over a long time. Typographers, print designers, information architects, interface designers and accessibility specialists worked out what makes a page readable and usable, and wrote it down, taught it and refined it.

We think it is worth being clear about this, partly out of respect, and partly because understanding where good design comes from makes you better at directing an AI tool.

The patterns are older than the web

Many of the rules that make a website feel calm and orderly come from print. Long before screens, typesetters knew that very long lines of text are tiring to read, that generous margins make a page feel considered, and that a limited set of type sizes creates a clear hierarchy. Designers working in the mid twentieth century, particularly in the Swiss tradition, developed rigorous grid systems that divided a page into columns and used them to align everything. If you have ever noticed that a well-designed site feels "lined up", you are seeing that inheritance.

The web then added its own layer. Over the years, designers and developers settled on conventions that users now rely on without thinking:

• The site name or logo sits at the top and usually takes you home.
• Navigation lives along the top or behind a clearly marked menu on small screens.
• Links look different from body text.
• Form fields have labels that stay visible.
• Important actions are styled as buttons, with one primary action per screen.

None of these is a law of nature. They are agreements, built up by countless people making sites, watching how visitors used them and adjusting. Their value comes from being shared. When a site follows them, visitors don't have to learn anything new.

What AI actually contributes

An AI design tool is, in a sense, a very fast and very consistent apprentice. It has absorbed these patterns and can combine them for a particular purpose: a restaurant, a portfolio, a community group, a consultancy. Its real contribution is threefold.

Selection. It can pick which patterns suit your situation. A photographer's portfolio wants large images and very little text; a solicitor's site wants clear services, trust signals and an easy way to get in touch.

Composition. It can arrange those patterns into a coherent page, with consistent spacing, type and colour throughout.

Speed. It can do all of this in moments, and redo it as often as you like.

That is valuable. But notice that the quality of the result depends almost entirely on the quality of the patterns underneath. If a tool is built on careless defaults, it will produce careless sites very efficiently. If it is built on the accumulated work of skilled people, it can produce something genuinely good.

Why "new" is not always better

There is a temptation, with any creative tool, to chase novelty. Unusual navigation, unexpected scrolling, a layout nobody has seen before. Sometimes that is exactly right, and talented designers do it brilliantly, with a clear reason and careful testing.

For most everyday websites, though, novelty is a cost. A visitor who arrives at a florist's site wants to see the flowers, check prices and order. Every unfamiliar interaction slows them down. Established patterns are not a failure of imagination; they are a kindness to the people using the site.

This is one place where AI and human craft line up neatly. The AI is good at the familiar, and for the majority of small sites, the familiar done well is what works.

How to use this when you design with AI

Knowing that the tool is drawing on human patterns gives you better ways to ask for things.

Name the pattern, not just the feeling

"Make it look better" gives the tool very little to work with. Try describing the pattern you want:

• "Use a single-column layout for the about page so it reads like an article."
• "Put the three services side by side as cards, each with a short heading and one line of description."
• "Add a clear call to action button at the end of each section."
• "Keep the navigation to four items."

Borrow vocabulary from designers

A few terms go a long way: hierarchy (what is most important), white space (room around things), contrast (difference in size, weight or colour), alignment (things lining up), rhythm (consistent spacing that repeats). Using these words in a prompt tends to produce more deliberate results.

Look at sites you admire and describe their structure

Not their branding, which belongs to them, but their structure. "A large headline with a short paragraph and one button, then a row of three examples, then testimonials." Structure is the shared language of the web; copying someone's identity is not.

Check the result against the basics

When a draft appears, ask a few questions that any trained designer would:

1. Is it obvious what the most important thing on the page is?
2. Are there more than three or four text sizes? If so, can some be merged?
3. Do things line up along clear edges?
4. Is the body text comfortable to read, without lines stretching across a wide screen?
5. Is there one main action, clearly styled?

If the answer to any of these is no, say so in plain words and ask for a change.

Respect where it is due

At Katzero we think about our tool as standing on a very large foundation laid by other people. Typographers who spent careers on legibility, designers who worked out grids by hand, developers and accessibility practitioners who documented how to make pages usable for everyone. An AI design tool is only as good as its respect for that work.

That also means being honest: when a project needs fresh thinking rather than well-applied patterns, a human designer is the right person to call. The patterns are a foundation. Building something truly new on top of them is still a human craft.