Explore. Build. Grow
Explore. Build. Grow

Our flexible three-stage framework to craft digital experiences.

Learn more
Expertise
Outcomes
View our portfolio
View our portfolio

Explore our range of digital solutions.

Learn more
Industries
Our Approach to AI
Our Approach to AI

Keeping ourselves at the forefront of emerging technology is part of our DNA. But we recognise that ‘technology for technology’s sake’ is not the end goal – there is a balance to be struck.

Learn more
DXP
Headless
Cloud Platforms
Marketing Optimisation
Contact Us
Contact Us
Explore. Build. Grow
Explore. Build. Grow

Our flexible three-stage framework to craft digital experiences.

Learn more
Expertise
Outcomes
View our portfolio
View our portfolio

Explore our range of digital solutions.

Learn more
Industries
Our Approach to AI
Our Approach to AI

Keeping ourselves at the forefront of emerging technology is part of our DNA. But we recognise that ‘technology for technology’s sake’ is not the end goal – there is a balance to be struck.

Learn more
DXP
Headless
Cloud Platforms
Marketing Optimisation
Contact Us
Female developer typing with decision nodes coming out of the laptop
Home / Insights / ＿

Will AI build your next website?

AI really can build a website in a few minutes. It just won't be your website.

Adam Griffith
Adam Griffith

22 September 2026

8 minute read

You've seen the demo. A sentence goes into a text box, 'build me a website for my company', and a few minutes later a website comes out. Hero image, tidy navigation, three trust logos, a contact form. The room does that little intake of breath, and somebody asks what this means for agencies. (We're a digital agency. We get asked a lot.)

So, will AI build your next website? Partly, yes, and the 'partly' is genuinely impressive. But not from one prompt. What happens between that prompt and a website worth having is the whole job, and most of it is decisions.

The tools are real

Lovable turns a plain-English prompt into a working application. It passed US$500 million in annualised revenue in June, on 8 million users and 146 staff, and by its own count 80 percent of the people building on it have no technical background. Vercel's v0 produces polished front end code from a description. Bolt lets you prompt, build and deploy a full-stack app without leaving the browser. And Ploy, founded by Webflow co-founder Bryant Chou and launched in June with US$27 million from First Round Capital and Y Combinator, deploys agents that keep working on your site after launch, watching rankings and competitors while you sleep.

We've also been on the journey to 'agentify' our delivery model. Our teams use these categories of tools every day, and the productivity gains are real. Anyone in our industry telling you it's all hype is either not paying attention or quietly terrified.

And yet.

What was in that prompt?

Think about what actually goes into a website that performs. Strategy, customer research, brand, information architecture, content, design, analytics, SEO and, increasingly, GEO, the craft of earning citations in AI answers as well as rankings in search. Then the technical layer, which is architecture, integrations, security and infrastructure.

That one-sentence prompt skipped none of them. Every one got decided, positioning, audience, structure, tone, stack, in the time it took to make a coffee, and almost none of it was put to you, your customers or your market.

Decided by what? Defaults. To be fair, they are not crude ones. Ask for a law firm and you get the law-firm average rather than some generic mush, because the model has read thousands of law firm websites and knows the shape of a good one. The defaults are industry-specific, and they are getting sharper. They are still defaults, which by definition is what everyone else in your industry is getting too.

A table showing all the different decision points involved in making a website.

A single prompt skips none of these. It answers all twelve at once, from defaults, and almost none of it is put to you, your customers or your market.

But doesn't it know best practice?

Yes. Better than most people assume, and often better than a rushed human. It has read every best practice guide ever written. If best practice were the whole job, this article would be over.

The problem is that most of those disciplines aren't bodies of knowledge at all. They're processes for manufacturing information that doesn't exist yet. What your actual customers actually need, the constraints peculiar to your sector, the content trapped in six people's heads and three legacy systems, the brand calls nobody has made. Best practice can tell you how to run a content audit. It cannot tell you what your content should say.

Hand that same one-line brief to any decent agency in the country and nobody would build from it. The good ones would ask questions, a few hundred of them, over weeks, of you, your data and your customers, before a pixel got pushed. AI rarely asks, and if it does, it's no more than a few quick clarifying questions. Then it answers, instantly, confidently, and with no idea whether any of it is right for you.

The problem isn't the intelligence. It's the obedience.

A website is a thousand decisions

Every website we build is the end point of thousands of decisions, large and small. Who this is for. What goes on the homepage. What gets cut. What counts as proof. What the primary navigation is called. Whether the case studies sit above the pricing or below it. Most of those calls are made by experts, with input from the client, and with a lot of context sitting behind them. Nobody types a website into existence. You decide your way to one, a few thousand times over.

AI is very good at the distance between those decisions. It closes the gap, and the quality of the work it produces along the way keeps improving. What it does not do is remove a single decision point. The tree is exactly as wide as it always was.

It can make the decisions too, and often make them well, but only with the right context in front of it. Without that context it still decides, because it always decides. It just decides the way everyone else does.

A diagram illustrating how a website can miss the mark by following 'default' decision processes, rather than tailoring the decision making to the specific needs of the brand.

Take a handful of those decisions differently and you finish somewhere else. Not slightly, either, and this is the part that surprises people. Ten straight yes-or-no decisions give you 1024 possible websites. Twenty give you just over a million. A hundred give you this many:

1,267,650,600,228,229,401,496,703,205,376

That is more possible websites than there are atoms in a human body. And it assumes every decision is a simple yes or no, when almost none of them are. Exactly one of those endings is the website your business actually needed.

The context nobody has written down

So the whole thing turns on context, and the context that matters most is the context nobody has written down.

It is in someone's head. It is the thing the GM of Retail will not say in a workshop but will say in a corridor afterwards. It is the product line that is quietly being wound down, the board member with strong views on the logo, the reason the last agency was let go. It is the support inbox nobody has read in two years, and the four sales people who each describe the business differently.

None of that is in the brief, because a brief is never the full picture, however good the client is. We routinely learn things in week 10 that nobody knew in week one. It surfaces because a person was in the room asking, noticing the pause before someone answered, and going back a second time. That is elicitation rather than retrieval, and no amount of plumbing gets you there.

Feed that context into the work and the decisions change. Every one of them, all the way down the tree.

A very nice display home

Back in 2022 I wrote about laying the right foundation for your digital asset, and made the point that small decisions can have big consequences down the track. The advice then was to call in an architect before anyone starts digging, and to measure twice so you only cut once. Four years and one AI revolution later, I would not change a word of it.

The one-prompt website is the display home. It photographs beautifully. The styling is sharp, the proportions are right, there are tasteful books on the shelf. And nobody lives there. The books are props, the kitchen has never cooked a meal, and it wasn't built on your block, with your slope and your soil and your council overlays.

Sometimes a display home is exactly what you need. A campaign landing page, a prototype to test before committing real money, an internal tool four people will use. For those jobs, one prompt into Claude, Lovable or Bolt is a brilliant answer, and pretending otherwise is snobbery. An enterprise website though is the most complicated house on the street. It has to carry a brand, convert considered purchases, integrate with a CRM and a DAM, and satisfy legal, accessibility and security teams for years.

Who made the decisions?

Bryant Chou from Ploy makes a point worth repeating, given he is building the thing that is meant to replace agencies: “When answer engines summarise your content and agents browse on your customers’ behalf, your website is the one thing you fully own.”

Websites matter more in this new world, not less.

Which raises the stakes on getting it right. AI has not made the build cheap. An enterprise website is still a serious piece of work, with a long tail of custom paths, integrations and edge cases that nothing generates for you. What AI has done is shorten the run between one decision and the next, and lift the quality of the work produced along the way. Our teams move faster and the output is better. The decisions are still ours together to make.

AI augments a team of experts. There is still a human at the wheel, working with your people, asking the questions, listening for the things nobody thought to write down, and making the calls that land you on the one-website-in-a-million that fits you and your customers.

So, will AI build your next website? More than partly. It will draft, design, code, test and migrate its way through much of the work. But the disciplines that make a website worth having don't get skipped just because nobody mentioned them in the prompt. They get answered, either by defaults, or by people who know your customers, your market and your numbers.

AI can absolutely build you a website. It just won't be yours. That one is at the end of a few thousand decisions, and every one of them starts with a question a human thought to ask.

Adam Griffith
Adam Griffith

22 September 2026

8 minute read

Enjoyed this article? Share it...

Want to tap into the expertise of an agency that’s been in operation since 1999?

Get in touch

Keep Reading

Want more? Here are some other blog posts you might be interested in.

Follow
Contact Us
1300 554 891
[email protected]

Our locations

Stay informed
Our best articles, observations and news delivered to your inbox
Subscribe
Flags

We acknowledge the Traditional Custodians of the land on which we work, live and engage. We pay respect to Elders past and present.

Luminary is a registered trademark of Luminary Digital Pty Ltd

Terms of Services
Privacy Policy