AI Built Websites: What It Actually Costs You After Launch

The current sales proposition is quite compelling. AI can create your website quickly and, in some cases, you can see the finished design before deciding whether you want to pay for it at all. If you are comparing that with a conventional website project involving planning, design, development, testing and ongoing conversations with an agency, I can understand why the AI option gets attention.

There is nothing wrong with asking why a website should still cost thousands of dollars if software can produce something attractive in a fraction of the time. I think that is a fair question.

The problem is that the preview shows you the easiest part to judge. You can see the design, read the words, click the buttons and decide whether you like the photographs. What you cannot easily see is:

• How the website has been structured underneath.
• Who controls the code and hosting.
• Whether another developer could take it over.
• How easy it will be to change.
• Whether accessibility has been properly considered.
• Whether the site gives you enough room to keep building your SEO and AI search visibility.

Those things tend to become much more important later.

AI has changed how quickly some websites can be produced. There is no point pretending otherwise. If AI allows somebody to build a website faster, the production cost may well be lower.

Where I think business owners need to be a little more careful is assuming that a lower build cost automatically means a lower website cost. They are not necessarily the same thing.

One of the things we have noticed when looking at AI built websites is how unremarkable the traffic result has sometimes been. I mean that literally. We have not seen some dramatic collapse across the sites we reviewed, but nor did we see a remarkable increase. Traffic was largely static.

That does not prove AI built websites cannot perform well. Of course they can. It is simply what we measured on the sites that came across our desk. But if the reason for replacing your website was to improve visibility, increase enquiries or create some form of commercial movement, a static result has not given you any of it.

This is where a very cheap website can stop feeling quite so cheap. You might ask for what appears to be a simple change. Add a service. Change part of the homepage. Create another location page. Move something in the navigation.

Before the new developer can make that change, they may first need to understand:

• Where that section is controlled.
• What else is connected to it.
• Why it was built that way.
• What could break if they change it.

You can end up paying somebody to answer those questions before they have even started the job you called them about. We have reached situations where repairing and reworking what is there becomes less commercially sensible than rebuilding it properly.

I would not suggest that happens with every AI built website. It does not. But it happens often enough for me to think the question belongs in the buying decision. What does the website cost to build, and what is it going to be like to own?

Diagram 2: The cost that does not appear in the website quote.

This sounds like such a simple question, but it is also one of the easiest to forget when somebody is showing you a finished website.

Imagine you launch today. Three months later you add a new service. Can you add the page yourself? A customer asks the same question for the tenth time and you decide it would make a useful article. Can you publish it? You want to update a photograph, add a new staff member, change some wording or alter your menu. Can you?

On a managed content system, many normal website changes can be handled without touching code. That does not mean there is never a need for a developer. There will be. But changing two paragraphs should not automatically become a development project.

With some generated sites, businesses only discover the limitation when they try to make the first meaningful change.

SEO is rarely one enormous piece of work. A lot of progress comes from small changes made consistently over time. You may want to:

• Expand a page that has started to rank.
• Add a useful internal link.
• Improve the title of a page.
• Answer another customer question.
• Create a page for a new service.
• Update an old article.
• Improve a heading.
• Add structured information.
• Fix a redirect.

None of those jobs sounds particularly large. If each one needs to be scoped, quoted, coded and deployed, businesses naturally make fewer changes. And eventually, they stop asking.

The site still works. Nothing obviously breaks. It just freezes. A year later, the business has changed but the website has not.

There is another risk that has nothing to do with whether the new design looks better. If an AI built website is replacing an existing site, the old URLs may already carry search history, links and rankings. Simply publishing a completely new structure over the top can throw some of that value away.

The old addresses should be mapped to the most relevant new pages and redirected properly. I would ask for that plan before launch, not after traffic drops. It is a fairly ordinary SEO job, but missing it can create an immediate problem for a site that was already visible in search.

Basic SEO helps make your website technically understandable. Deeper SEO gives search engines and AI systems more evidence about why your business should be trusted on the subject.

Deeper SEO is not about creating hundreds of pages simply because you can. In fact, that can take you in completely the wrong direction. It is about properly developing the subjects that sit around what your business does.

Take a plumbing company. A page saying you offer drainage services is useful, but customers have questions beyond the word drainage. They may want to know:

• Why the same drain keeps blocking.
• When CCTV inspection makes sense.
• What the signs of a damaged underground pipe are.
• Whether tree roots could be involved.
• Whether drainage issues differ between older and newer properties.

Now the website starts demonstrating knowledge around the service instead of simply stating that the service exists. The same applies in almost every industry.

Your services connect to real customer questions. Those questions connect to examples, experience, locations, people and supporting knowledge. Internal links connect related information. Reviews and mentions outside your website add another layer. Other sites may refer to your business. Named people become associated with particular expertise.

That starts to create a much clearer picture of who you are and what you actually know.

One of the questions I hear more often now is, how do we get mentioned by AI? There is no guaranteed formula. I would be very cautious of somebody who tells you there is.

What we can work on is the information and evidence AI systems have available when they are trying to understand your business or answer a question related to your field.

Google’s May 2026 guidance confirms that normal SEO foundations remain relevant for its own generative AI features. It also recommends expert led, useful information that provides something beyond common knowledge. [2]

So I do not see AI visibility as something sitting in a completely separate box from SEO. For me, deeper SEO helps build the foundation for it.

You still need a technically sound site and search systems still need to access and understand the pages. But authority and trust require more. That can include:

• Real expertise.
• Connected subject coverage.
• First hand information.
• Useful examples.
• Up to date information.
• Mentions and references elsewhere online.
• Reviews
• Links
• Consistent business information.
• Named people associated with what the business knows.

You cannot simply tell an AI system that you are an authority and expect it to accept the claim. You need to give it evidence.

This may be the quieter cost of a website that is difficult to edit. Nothing dramatic happens. You simply stop adding things.

A useful question comes up and nobody publishes the answer. A service changes and the page stays the same. A page starts appearing in search, but nobody develops it further. An internal link would help someone move between two useful pieces of information, but it becomes another developer request.

Meanwhile, another business keeps building. They answer more questions, add examples, update older material and develop supporting pages around the services they want to be known for.

After two years, the difference between the two websites is no longer simply how they were originally built. One has accumulated knowledge. The other still looks much the same as it did on launch day.

This is why I put editability so high on the list of things a business owner should ask about. Can your website keep showing what your business knows? If it cannot, that becomes more than a website management problem. It can become an SEO and AI visibility problem too.

It is very easy to turn AI security into scare material. We have not personally reviewed one of these websites and found exposed database passwords or completely open administration areas. I am not going to suggest that we have.

There is still credible security research showing why generated code should be technically reviewed. Researchers presenting at the 2025 USENIX Security Symposium tested 16 code generating language models and produced 576,000 code samples. One of the issues they examined was package hallucination, where a model refers to a software package that does not actually exist. [3]

You are focused on running your trade or your service, and digital moves very quickly. I would therefore not expect you to understand the software dependency chain sitting behind your website. I would expect somebody involved in building it to understand:

• What the website relies on.
• What needs maintaining.
• What needs updating.
• What needs checking.

AI can produce code quickly. Somebody still needs to take responsibility for it.

Accessibility is another area where the finished website can look fine to one person and be difficult for somebody else to use. You may navigate with a mouse. Another visitor may use a keyboard. Someone else may rely on a screen reader.

Heading structure, form labels, navigation order, image descriptions and other details suddenly matter in a very different way.

There is research directly related to AI website builders here. Researchers Zsuzsanna Palmer and Sushil Oswal created business websites using three generative AI website builders and tested the finished sites using both accessibility tools and a screen reader. None of the three websites they tested was fully accessible to screen reader users. The researchers identified issues around layout, navigation, headings and other accessibility elements. [4]

It is a study of three tools. We should not turn that into a claim about every AI built website. But accessibility needs to be tested. A nice looking preview does not tell you whether the work has been done.

I would ask this before becoming too attached to a design. What exactly do you own?

If your website sits inside an AI website platform, you should know:

• Whether you can take it somewhere else.
• Whether the code can be exported.
• Whether another developer can realistically use that export.
• What happens if you stop paying for the platform.
• Who controls the hosting account.
• Whether you can hand the site to another developer tomorrow.

Ownership and portability are not always the same thing. You might own the words and images but still depend on the platform to keep the site working. That does not automatically make the platform a bad choice. You simply need to know before you commit.

I would also ask who holds the commercial rights to the images, fonts, code and other assets used on the site. That question is easy to miss when everything arrives as one finished package. It becomes a real problem if you later move platforms or hand the website to somebody else.

There is another part of an AI website build that the finished site will never show you. The prompts.

What information was entered while the site was being created? Perhaps nothing sensitive at all. But a website project can involve customer information, internal plans, staff details, pricing information and material that has not been published yet.

The New Zealand Office of the Privacy Commissioner advises organisations using generative AI to consider carefully what personal or confidential information is entered into these systems and how that information may be retained or disclosed. [5]

If personal information is disclosed to an organisation outside New Zealand, Information Privacy Principle 12 may also need to be considered, depending on the circumstances and how the overseas provider is handling that information. [6]

You do not need to become a privacy lawyer to commission a website. It is reasonable to ask the person building it what tools they are using and what information is being supplied to those tools.

You do not need to understand code to ask good questions. Before accepting the website, I would ask:

I would want clear answers. Not technical answers designed to make the questions disappear. Plain English answers.

You do not need to become a developer before buying a website. But I would make a few practical checks.

1. Ask whether all important page content can be properly accessed and indexed by search engines.
2. Confirm that the website has not accidentally been blocked from search through a noindex setting or another technical restriction.
3. If the website replaces an existing site, make sure important old URLs have been redirected to their new locations.
4. Check that important pages have sensible titles, descriptions and headings.
5. Test every important form on both a desktop computer and a real mobile phone.
6. Have somebody technically competent look at the website underneath the design, particularly if the business handles personal information, payments or user accounts.
7. Ask who maintains the website, when the backup was last successfully restored, and who is responsible for recovery if something goes wrong.\

One more question: Can you actually change it?

That question is surprisingly easy to forget when the website already looks finished.

Are AI Built Websites Bad for SEO?

Not automatically. An AI built website can have perfectly acceptable basic SEO. It can have good titles, headings, descriptions, sitemaps and crawlable pages.

The bigger question is whether the website can support deeper SEO over time. That means useful supporting content, strong internal relationships between pages, real expertise, ongoing updates and enough depth around your subject for search systems to understand what your business actually knows.

Can an AI Built Website Rank in Google?

Yes. Google does not exclude a website because AI helped build it.

Technical quality, useful information, page experience, relevance and the wider signals around the website still matter. Google has also confirmed that ordinary SEO foundations continue to underpin visibility within its generative AI search features. [2]

Can an AI Built Website Be Mentioned in AI Search?

Yes, but nobody can guarantee a mention.

If you want to increase the chances of your business being surfaced or referenced, you need to give AI search systems useful information to work with. That includes strong technical SEO, subject depth, first hand knowledge, connected content, external mentions, reviews and other supporting evidence around the business. This is why I see AI visibility as being built on deeper SEO rather than replacing it.

Is an AI Built Website Cheaper?

It may be cheaper to produce. That does not automatically mean it is cheaper to own.

Look at the cost of future changes, ongoing maintenance, hosting, platform fees, SEO work and what happens if another developer needs to take the site over. The initial build price only tells you what it costs to get started.

Can I Move an AI Built Website to Another Provider?

That depends on the platform and how the website has been created. Some sites can be exported. Others depend heavily on the platform they were created within.

Ask this before you buy. You want to know whether another developer can realistically take over the website rather than simply being told that you technically own the content.

Who Should Review an AI Built Website?

Someone who understands website development and can assess more than the finished design.

Depending on the website, that may include reviewing the code, hosting, security, forms, accessibility, search setup, redirects, analytics, backups and how future changes will be handled.

What Is the Difference Between Basic SEO and Deeper SEO?

Basic SEO makes sure search engines can properly access and understand the website.

Deeper SEO builds around the subjects you want your business to be known for. It brings in customer questions, expert knowledge, supporting pages, internal links, first hand experience, local context where relevant and external evidence such as mentions and reviews. That deeper layer helps build the authority and trust needed for stronger organic visibility and increasingly for visibility within AI search.

Cathy Mellett is Director of Net Branding Limited. Net Branding has worked across websites, SEO, paid media and digital marketing since 2008. Today, that work also includes AI search visibility and helping businesses understand how traditional SEO, authority, trust and AI search fit together.

We use AI within our own work. We also believe that somebody competent still needs to understand, review and take responsibility for what AI produces.

Net Branding Limited
165 Orakei Road, Remuera, Auckland 1050
Phone: 09 523 0478
Email: sales@netbranding.co.nz
Website: www.netbranding.co.nz