The most important part of a website isn't the code.

A website can be beautifully designed, technically sound, and completely wrong for the business it represents.

That is easier to do than most people realize. Ask an AI tool to create a website for a small business, and a few moments later it produces something polished. The layout is clean. The buttons work. The pages look right on a phone.

It could also belong to almost anyone. The language is broad, the services sound interchangeable, and the photos feel disconnected from the people behind the business.

The website functions, but it does not understand the business. That is the part people often miss. The most important part of a website is not the code. It is the understanding behind it.

The invisible work comes first.

Code matters. A website should load quickly, work on every screen, be usable by everyone, and keep visitors' information safe. Good code makes that possible.

But code can only build what it has been told to build. It cannot decide what matters most to a customer, or notice that a service name makes sense inside the company but confuses everyone outside it.

Before a useful website can be built, someone has to answer harder questions:

• What does this business actually do?
• Who is it trying to reach?
• What does that person need to understand before getting in touch?
• What makes the business meaningfully different?
• What does the owner know after years in the business that has never been written down?

Those decisions shape the website long before the first line of code is written.

Better tools do not remove the need for judgment.

AI has made it easier to produce code, layouts, images, and written content. That is useful. It is also why judgment matters more now, not less: when production gets faster, it gets easier to produce the wrong thing quickly.

An AI tool can write twenty headlines in a minute. That does not mean any of them understand the customer. The problem is rarely that AI cannot produce enough. It is that AI does not automatically know what deserves to be produced.

A weak process asks:

What can this tool make?

A stronger process asks:

What does this business need, and how can the tool help us make it well?

Restraint is part of the job.

Good website work is often measured by what gets removed, simplified, or declined. A homepage does not become more persuasive because it makes more claims, and a visitor does not feel more confident because five buttons compete for attention.

Sometimes a business needs a system it can edit itself, like WordPress. Sometimes it needs a lightweight custom-coded site, written from scratch, with fewer moving parts. Sometimes Squarespace is enough. Custom-built is our flagship, not our only answer.

The job is to understand how the business runs, who will maintain the website, and what the customer needs to do. Then the platform, meaning what the site is built on, can follow the decision, not the other way around.

A website needs a business brain.

Before I ask AI to help with a website, I want it to understand the business behind it. Not perfectly. Just enough that it isn't working from a generic request. I think of this collected knowledge as the business brain:

• The company's story, services, and values.
• Its customers and the area it serves.
• Its tone of voice and preferred words.
• Pricing, policies, and common questions.
• The claims it can back up, and the ones it should avoid.
• The decisions made along the way.

Most businesses already have this knowledge, but it is scattered across the current website, proposals, emails, and the owner's head. Owners know the questions customers ask before calling and the small details that build trust. Much of it feels so obvious that it never gets written down.

AI cannot use knowledge that has never been written down.

Organized knowledge changes everything.

Writing it down turns scattered knowledge into something you can use. It can start small: a few focused documents for your brand voice, your customers and services, and the decisions you've made. You can write them yourself in an afternoon or build them with your designer. Our guide to building an AI business brief (a short set of documents that tells AI tools about your business) walks through five of them. The point is to give each important fact one reliable home.

That matters because websites rarely become inconsistent all at once. A price changes on the service page but not in the behind-the-scenes labels search engines read. A new phrase appears on the homepage while older pages keep the old wording. Months later, the website contains several versions of the truth.

AI has the same problem. Hand it a pile of documents, old and new, and it can't tell which are current unless you say so. It may treat an old plan as today's plan, or repeat a temporary idea as a permanent rule. Not because the tool is careless, but because the context is unclear.

Nerd note How we keep a project's facts in order

On our own projects, every source of information has a rank, so an AI tool (or a person) always knows which one wins when two disagree:

• Current decisions outrank old ideas.
• The live website outranks outdated descriptions of it.
• Each topic has one document that owns its facts.
• Temporary notes stay temporary.
• Archived material is reference, not instruction.

This kind of structure is not glamorous, but it prevents expensive confusion.

Organized knowledge also changes what you can ask for. Without context, a request sounds like this:

Write a service page for a furniture restoration studio.

The result will probably sound like a hundred other studios. With context, the request changes:

Write the restoration service page using our approved tone, our actual services, what we don't offer, and our description of who we serve. Don't promise prices or completion dates, and don't make claims we can't back up.

Now the AI is not inventing the business. It is working within it.

Different tools. One responsibility.

I use different AI tools for different kinds of work, and the tools will keep changing. The responsibility does not. Every tool should work from the same understanding of the project, stay within an approved scope, and have its work reviewed by a person. A strong process, the one we follow and the one to look for in anyone you hire, looks like this:

1. Understand the business.
2. Write down what you know.
3. Decide what this piece of work should achieve.
4. Look at what already exists.
5. Make a plan.
6. Review the plan.
7. Build only what was approved.
8. Check the result.
9. Fix only what needs fixing.
10. Record the decisions for next time.

AI can help with every step. It should not erase any of them.

Nerd note The tools I actually use

ChatGPT is often where I explore ideas, look at a problem from several directions, question an assumption, or refine language. Claude Code is often where the approved direction meets the actual project: it can read the site's files, understand how they fit together, propose a plan, make controlled changes, and check the result.

Some standards carry from one project to the next: forms are protected, passwords and keys never appear in public files, images have set dimensions, pages work with a keyboard, and the hidden labels search engines read never claim something the page doesn't say. I keep those in reusable instructions, separate from each project's own facts. A reusable instruction may say that prices must come from one authoritative source; the project supplies the actual prices.

Experience still has to recognize what is wrong.

Documentation makes AI better informed. It does not make judgment automatic. Someone still has to notice when a headline sounds polished but says nothing, when the menu reflects how the company is organized instead of what customers ask, or when technically correct photography feels wrong for the brand.

Those decisions come from experience. My background in web design and commercial photography production has taught me that the visible result is the final layer of a much larger process. A finished photograph may look effortless, but behind it are decisions about the subject, lighting, composition, timing, color, retouching, and what stays outside the frame.

A website works the same way.

AI amplifies organized experience.

There is a common fear that AI will make experience less valuable. In my work, I have seen the opposite.

Put simply, AI makes whatever you give it bigger. Give it organized knowledge and you get better work. Give it confusion and you get more confusion.

If the business is poorly understood, AI produces more generic writing and more code that solves the wrong problem. If the business knowledge is organized, AI can check work against approved decisions, spot inconsistencies, draft in an established voice, and handle repetitive work without losing patience.

The improvement does not come from the AI alone. It comes from the tool, the context, the process, and the person responsible for the outcome, working together.

AI amplifies organized experience.

What a business owner should look for.

You don't need to follow all of this to hire well. What matters is whether a designer tries to understand your business before they start building, and their questions will tell you.

The code is where the decisions become real.

None of this makes code unimportant. A thoughtful website that loads slowly, breaks on a phone, exposes private information, or becomes impossible to maintain is not a successful website. The code carries the decisions into the real world. It just cannot supply them.

The strongest websites begin with understanding. Written down, that understanding becomes direction for the design, the writing, the photos, and the search setup. Then the code gives the whole thing form.

AI can generate a website quickly. Understanding the business behind it still takes time, and that is the part that shapes every decision that follows. The code simply gives those decisions a place to live.