AI ends the world? Website design view from NZ

People ask strange questions once AI turns up in the news every second day. One week it is writing school essays, the next it is making fake voices, then suddenly someone wants to know if AI ends the world and whether that means they should keep it well away from their business website.

My short answer is no, I do not think AI is going to end the world through website design. I do think it can absolutely wreck a website if you use it lazily, trust it too much, or let it replace the human bits that matter, the judgement, the structure, the local nuance, the boring maintenance work that actually keeps a site useful.

That is the part I wish got said more plainly. The danger is not some sci-fi robot deciding your plumbing site should overthrow civilisation. It is much duller than that, and because it is dull people miss it. The real risk is a business owner publishing generic AI copy, stock-standard layouts, wrong service information, made-up FAQs, and then wondering why the phone stays quiet.

We build, host and maintain websites for New Zealand businesses, a lot of them service businesses, owner-led companies, organisations that do not have time to babysit five separate suppliers. So I have a pretty practical view on this. If a tool helps us move faster without lowering the standard, good. If it introduces noise, risk or fakery, out it goes.

Key takeaways

AI is useful in website work when it does the rough, repetitive, or exploratory part, drafting alternatives, helping sort content themes, speeding up admin, spotting patterns in messy information, and leaving the final decisions to people who know the business and the customer. It is not safe when it is treated like an automatic truth machine, or when people use it to avoid thinking about positioning, page structure, search intent, accessibility, performance, and what a visitor actually needs to trust you.

If you run a small business in New Zealand, the practical question is not whether AI is good or evil. That is theatre. The useful question is where AI belongs in your website process, where it does not, and who is checking the work before it goes live. That is where outcomes split. Same tool, very different website.

Why this question keeps coming up

I think people are asking three different things at once and bundling them into one dramatic sentence.

First, they are asking whether AI is dangerous in the broad, civilisation-sized sense. I am not going to pretend a website studio has solved that one over lunch. Second, they are asking whether AI is safe for website design, meaning privacy, copyright, accuracy, brand damage, bad advice, that sort of thing. Third, and this is usually the real one, they are asking if using AI will make their website feel cheap, fake or invisible in Google.

That third question matters most to the businesses we talk to. Because the website is not a philosophy exercise. It has to load properly on a phone in a driveway, explain the service clearly, earn trust quickly, and give someone an easy next step. If AI helps that happen, fine. If it gets in the way, I do not care how clever the tool is.

My view: AI is overrated and underrated at the same time

That sounds like fence-sitting, but it is not.

AI is overrated when people talk like it can replace a proper discovery process, content strategy, page hierarchy, search intent research, technical SEO, image optimisation, conversion thinking, hosting decisions, and ongoing maintenance. It cannot. Or rather, it can imitate pieces of those tasks just enough to fool someone who does not do this work every day.

It is underrated when people dismiss it as a gimmick. Used properly, AI can shorten the ugly first-draft phase, help break a blank page, surface repeated customer questions hidden in messy notes, and save time on support tasks that do not need handcrafted genius. That matters, because time saved in the right place can be spent on the parts that actually change whether a website performs.

And that is my bias, I suppose. I do not care whether a process sounds futuristic. I care whether the finished site is fast, clear, easy to maintain, and more likely to generate enquiries six months from now.

Where AI use for websites actually helps

There are parts of website work that are repetitive in a useful way. Pattern recognition, rough drafting, reformatting, summarising notes, helping compare headline options, turning a long ramble from a client into a cleaner first pass, these are fair jobs for AI. Not glamorous ones either, but real.

Sometimes a business owner knows their trade inside out and still struggles to explain it on a page. That is normal. Good website copy is not just expertise dumped onto a screen. It needs order. It needs emphasis. It needs a sentence that says the plain thing first. AI can help us get from the messy version to a draft we can shape.

It can also help with support work around FAQs, schema planning, metadata options, and identifying the obvious content gaps. I am careful there, because AI loves sounding certain while being wrong, but as a thinking partner or rough sorting tool it can be handy.

Same with ideation. If we are looking at ten possible ways to frame a service, or trying to stress-test whether a page is too broad, AI can speed up the early comparison stage. Not the final call. The early stage.

That distinction matters more than people think.

Where I would not trust AI on its own

I would not trust AI alone to write a service page for a local business and publish it untouched. Not if the business wants leads, not if they care about sounding credible, and definitely not if the service has any technical, legal, health, financial or safety edge to it.

The failures are often boring. Wrong terminology. Repeated phrases. Claims that sound puffed up. Explanations that are technically fine but useless to an actual customer. Location language that feels imported. A weirdly American tone. Advice that is broad when the page needs specifics. Or the opposite, fake specifics that were never true in the first place.

Then there is structure. AI often gives you something that looks complete while quietly missing the one thing the page needed. No proof path, no clear service area mention, no next step, no objection handling, no sense of what makes this business the obvious call. It looks polished enough to pass a skim, which is why people publish it too early.

That is why I keep saying the problem is not the tool, it is the trust you place in the first output. Actually no, let me correct that. Sometimes the problem is the tool, but only when people use a tool built for speed and novelty as if it were built for accuracy and accountability.

Is AI safe for website design? Safer than panic, less safe than caution

If someone asks me, “is AI safe for website design?”, I do not answer with a yes or no. I ask, safe in what sense?

Safe for productivity, usually yes. Safe for privacy, maybe, depending on what you feed it. Safe for factual accuracy, not by default. Safe for brand reputation, only if a human with taste and context reviews the output. Safe for SEO, again, only if the content is original in thought, useful, and actually aligned to what searchers need.

A lot of the panic comes from treating AI as one thing. It is not. There is a difference between using it to generate five rough headline directions and using it to produce your entire website copy, FAQs, service descriptions, testimonials style, image prompts and blog strategy with no review. Those are not morally equivalent. They are not even professionally close.

The safety issue I care about most is hidden sloppiness. If AI helps a team move fast, good. If it lowers the friction to publishing rubbish, that is dangerous in the very ordinary business sense. You lose trust, rankings wobble, enquiries drop, and then someone says “AI does not work” when the real issue was unedited output and weak judgement.

The boring risks nobody mentions enough

This is the bit that decides outcomes, and it is never the flashy part.

One risk is feeding confidential business information into tools without thinking through where that data goes. Another is publishing copy that sounds authoritative but contains subtle errors the owner does not spot because, fair enough, they are busy running the business. Then there is copyright and originality around imagery and design references, which gets muddy quickly if people start lifting styles too literally or relying on generated visuals where real photography would build far more trust.

There is also the maintenance problem. AI can help produce a site faster, but it does not maintain plugin compatibility, check form delivery issues, monitor uptime, test mobile layouts after updates, or notice that the booking button vanished on one device size. A website still needs grown-up care. I know that sounds unexciting. It is also true.

And one more thing. AI tends to flatten difference. If ten businesses in the same industry ask similar prompts, they get similar copy patterns back. Similar headings, similar claims, similar FAQs, similar vague promises about quality and service. That is not just a style issue. It becomes a search and conversion issue because your site stops giving people a reason to remember you.

What small business owners get wrong when they try AI alone

Usually they start with the copy before they know the job of the page.

I see this a lot. A business owner opens an AI tool and asks for a homepage, or for SEO text, or for “something professional”. The output comes back tidy-looking, and because it is less painful than starting from nothing, they keep going. But the page has no real brief behind it. No target search intent, no service priority, no conversion goal, no understanding of what the reader is worried about, no local context, no hierarchy that says this matters first and that can wait.

So they end up polishing language on a page that was strategically wrong from line one.

The other mistake is trying to save money in the place that costs most later. They use AI to replace planning, then spend months with a site that technically exists but does not really work. Slow pages, vague copy, duplicated ideas, odd metadata, weak calls to action, and too many plugins because no one thought through the build cleanly. None of that feels dramatic on launch day. It shows up later, when the site is hard to update and not pulling its weight.

What we actually use AI for, and what we still do ourselves

We use AI as a helper, not a decision-maker.

That means it can support early drafting, help us compare wording angles, assist with content organisation, and speed up some research and admin work. It can be useful for turning rough notes into something we can interrogate properly. It can help spot where a FAQ is missing or where a paragraph has gone muddy. For some maintenance and support tasks, it can also speed up troubleshooting paths or documentation.

What we do not outsource to AI is the judgement that makes a website perform. We still decide what the page is for, what should be said first, what should be cut, what sounds trustworthy in New Zealand English, how the site should be structured, how to balance design with speed, what hosting setup makes sense, what technical SEO needs handling from the start, and whether the final page would persuade a real person instead of merely sounding complete.

That last point is easy to miss. A website can look finished and still not be persuasive. AI is quite good at finished-looking.

If you want the whole thing done properly, design, hosting, maintenance, SEO, the practical day-two support that keeps a site from drifting, have a look at https://nzdh.co.nz. We built our process around that joined-up reality because most small businesses do not need more fragmentation. They need one team that can think and then keep thinking after launch.

AI and SEO: the wrong question is “will Google punish it?”

I hear this one all the time. “Will Google penalise AI content?”

That is not the first question I would ask. The better question is whether the page is useful, specific, accurate, original in its thinking, and satisfying for the searcher. A lot of AI content fails there long before any algorithmic theory matters. It is not that the page is AI, it is that the page is thin, repetitive, generic, or obviously written without firsthand understanding.

Search visibility has always had this awkward truth in it: machines can measure some things, but users reveal the rest. If the page does not answer the actual question, if it hides the useful detail under waffle, if it sounds like every other service page in the country, people bounce or skim or keep searching. That is the failure.

I would also say this plainly: AI often encourages overproduction. Too many pages, too many blogs, too much filler, all published because it was easy. For a small business site, I would rather have fewer pages that are genuinely strong than a content graveyard full of machine-assisted mush.

We build SEO and AEO in from the start because retrofitting it later is messy and usually more expensive in attention if not in dollars. If you want to see how we approach that in practice, not as a buzzword stack, start here https://nzdh.co.nz.

AEO and AI-ready websites are not the same as AI-written websites

This point gets blurred constantly, and it causes confusion.

An AI-ready website, or a site built with answer engine optimisation in mind, is a website structured so machines can understand it and people can still use it easily. Clear headings, sensible page hierarchy, direct answers, strong service pages, coherent internal linking, technically sound markup, and content that says one thing clearly instead of ten things vaguely. That does not require the copy itself to be machine-written.

In fact, the strongest AEO work often comes from very human writing. Short honest answers, direct language, specific service detail, and FAQs that sound like what customers actually ask on the phone. AI can help surface patterns in those questions, sure. But the final answer usually needs a person who knows where customers get confused and what wording calms them down.

That is one of the small tells I watch for. If a FAQ sounds polished but not speakable, if no real customer would phrase it that way, then something has gone off. Websites are full of that now.

Design work is not just words, and AI struggles there more than people admit

Anyone can generate a pretty layout concept now. That is not the same thing as designing a working website.

Real website design lives in constraints. What happens on mobile. What gets seen first. How much copy a section can carry before it breaks. Whether a button label makes sense. Whether the visual hierarchy supports the service priority. Whether the page loads quickly enough. Whether the images help trust or just decorate. Whether the footer quietly does useful work. AI can imitate patterns from existing sites, but it does not naturally care about these trade-offs unless a human keeps dragging it back to the actual job.

And then there is sameness again. A lot of AI-assisted design work has a certain look to it, smooth gradients, familiar block structures, polished but personality-light. Fine for a placeholder. Weak for a business trying to be remembered.

I am not anti-polish. I am anti-generic.

How I would decide whether AI belongs in your website project

I would start with the stakes.

If the site is a small brochure site and you need help getting rough ideas out of your head, AI can be a useful support tool. If the site handles sensitive information, carries technical claims, or sits in a competitive search space where trust and accuracy matter a lot, then the review process needs to be tighter and the human involvement stronger. That should not be controversial, but somehow it still is.

Then I would look at the maturity of your business message. If you do not yet know what makes you different, AI will not solve that. It will camouflage the problem for a while. Better to do the harder thinking first. What do customers ask, what do they hesitate over, what do they compare you against, what service pages actually deserve prominence, what proof can you show without overclaiming. Once that is clearer, AI can help accelerate parts of the execution.

And finally, I would look at who is accountable. If no one with real judgement is reviewing the output, then no, I would not call that safe. Fast, maybe. Safe, no.

So, is AI going to end the world? Not from your website. But it can still waste your time

I do not think your next homepage draft is the frontier of human extinction. I do think AI gives people a false sense that the hard work has been done when only the typing has been done.

That distinction matters because websites fail quietly. They do not always crash. They just underperform. They confuse. They blend in. They ask too much reading from the wrong visitor. They answer the wrong question. They look fine in a portfolio and do very little in the wild. AI can produce that kind of failure at speed.

Used carefully, though, it is just another tool, and a decent one. It can help us move faster on the repetitive parts and spend more energy where websites actually win, clarity, structure, trust, speed, search visibility, and maintenance that does not get neglected after launch.

That is where I land. No panic, no worship. Use AI for websites where it genuinely helps, keep humans in charge of the important calls, and do not confuse easy output with good work.

FAQs