Why Good Websites Have Not Got Cheaper, Even Though AI "Builds Sites for Free"

You have two things in front of you at once: a quote for €3,200-6,000 and an ad for a builder promising an AI website in three minutes, for free. Both are true. That is the paradox anyone collecting offers in 2026 runs into: if AI generates websites from a single prompt, why does a website still cost that much?

Before I answer, a fair disclaimer. I build websites for a living, so I have an interest in prices staying where they are, and you have every right to read this with suspicion. Let me add a second fact though: I use AI every day. Claude (including Claude Code for writing code), ChatGPT, Perplexity - these are standing tools of my work, not a curiosity. And my quotes still have not dropped. This post explains why, and it starts with a list of the things that genuinely did get cheaper.

The short answer (for the impatient)

AI sped up writing code by roughly a third, but that saving did not turn into a discount - it turned into quality: safer, faster sites that are easier to extend, at the same price. What did not get cheaper: responsibility for the site working, fitting it to your specific business, copy that sells, maintenance and security, or the time it takes to understand how your company makes money. The price of a website drops exactly where the contractor's responsibility drops - a free website means nobody is answerable for it.

If you need it in one sentence: when a website only has to exist, an AI builder is enough; when it has to earn, you are paying for the things AI did not do and will not do for a long time.

What actually did get cheaper (yes, AI works)

Let me start with the part you will not hear from contractors defending their prices: some of my work genuinely did speed up.

AI handles layout drafts, the code of repetitive elements, graphic variants, first drafts of copy and research well. Across my projects, time spent writing code went down by about 30 percent - with the caveat that it does not work out that way in every project. There are tasks where AI saves hours, and tasks where fixing generated code costs more than writing it from scratch would have.

So why did the quote not drop by those 30 percent? Because the saving went somewhere else. My prices through 2025 and 2026 have held, but for the same amount a client gets more today: safer, faster, more reusable components. "Reusable" simply means a section built properly once comes back on later pages and future extensions for a fraction of the original cost - and you see that difference not in the launch quote, but in the quote for next year's expansion. A discount is something you would consume once. A better component works for you for years: fewer failures, cheaper development.

That is the first half of the answer: the gain from AI went into quality, not into a price cut. The second half is more interesting - the rest of the cost did not fall at all.

What the price of a website is really made of

In my projects, writing code itself is roughly 70-75 percent of the time. I stress roughly, because it depends heavily on the client and the requirements. In projects with business logic inside them, a calculator or a configurator for instance, the split can fall to 50-50 - because before the first line of code exists, the conversations have to map out how the company actually operates. AI will not shorten a conversation whose whole point is to find out something nobody has written down anywhere yet.

And here is the crux. Code, the largest slice of the time, sped up by a third. Everything else did not speed up by a single minute:

• Understanding the business. AI does not know who your customers are, what you actually make money on, or how your sales process runs. A prompt does not replace an interview - and a website that has no answers to those questions is a pretty shell.
• Decisions. AI will generate "a" website instantly. The right website comes out of a series of decisions: what goes on it, in what order, what gets cut. Someone has to make those calls and be able to justify them.
• Copy that sells. AI writes correctly and averagely. Text written for one specific company and one specific search phrase is still work - which is exactly why sites generated in bulk look and sound the same.
• Responsibility. The warranty, the fixes, the person who picks up the phone when the site goes down on a Friday afternoon.
• Maintenance. Updates, backups, security. With me, ongoing technical care runs €320-600 per month, and that line item has nothing to do with how fast the code was written.

AI sped up production, but it touched neither responsibility nor the understanding of your business. You are paying for a website that fits the company and that someone is answerable for - and those two things cannot be generated. What those components look like on a concrete product, I broke down in how much a WooCommerce store costs.

The price drops exactly where the responsibility drops

If the work did not disappear, where does "a free website" come from? The mechanism is simple: in an AI builder the work does not vanish, it changes owner. You are now the webmaster, the editor and the administrator. And whatever you do not take on turns into a subscription - in real use, meaning your own domain and no builder ads, price lists start at roughly €20-40 net per month, as this breakdown of AI website costs shows. The generator is free. A working website is not.

The easiest way to see it is by analogy: a free dinner exists, and it is called "cook it yourself". Nobody claims restaurants are lying about their prices because recipes exist online. An AI builder handed you the recipe and the pan - that is genuinely a lot, and genuinely free. It did not hand you a chef, or anyone answerable for the dish turning out.

And to be clear: this is an honest deal if you know what you are taking on. The problem starts in two places. When someone sells you an "AI website" at the price of one built by hand. Or when a free website costs you customers you never find out existed.

The two-year cost of a cheap website

The price of a website is worth counting over time, not on launch day - I laid that method out in detail in website template vs custom theme, and it works identically here.

On the builder side you have the subscription (€240-480 net per year on entry-level plans), your own time for every change, and limits you only hit once they hurt: payment or booking integrations, what you can do for SEO, and above all no export - a site built in a builder usually cannot be taken with you. It is not yours, you are renting it. On the custom-built side you have a higher entry cost and then maintenance whose cost categories are visible and predictable: hosting, domain, updates, care.

The most expensive line, though, is hidden at the end: the cost of moving. When a company outgrows the builder, the migration is practically always a rebuild from zero, with a new structure, new SEO and copy to rewrite. I collected the full set of strengths and weaknesses of that kind of start in pros and cons of an AI-generated website.

When a free or cheap AI website really is enough

There is no catch in this section, because these are the situations where I talk people out of spending €3,200:

• You are testing a business idea. The site is there to verify demand, not to build a brand. An AI builder is the cheapest possible experiment - and that is exactly what it is for.
• You need a business card, "just so it exists". Customers come by referral and the site only has to confirm you are real. A decent builder page does that for a fraction of the price.
• Your budget is below roughly €800. An honest builder page beats a "custom" website for €600 that is not custom at all.
• You operate seasonally or one-off. An event, an intake, a campaign - a three-month site does not need foundations built for a decade.

If a year from now that site starts earning and you outgrow it, you will come back to the subject deliberately. That is a good scenario, not a failure.

How to read a quote so you know what you are paying for

A story from my practice. A client was frightened by the quote at first. They accepted it after receiving the offer broken down in detail - including hours set aside for conversations about features, so those features would fit their business exactly. The price did not change. What changed was that they could see what was inside it. A quote you are afraid of is usually a quote you do not understand.

You can see what a broken-down offer looks like in my sample quote for an online store - a complete scope of work split into stages, with a separate section on what pushes the price up and what pulls it down. It deliberately carries no amounts: the same scope costs differently in two different stores.

So instead of comparing amounts, check five things:

1. Ask for the quote broken down into stages. A contractor who cannot break down their own price is quoting by gut feel.
2. Ask what happens after launch and what care costs. No answer means a hidden cost you will discover at the worst moment.
3. Ask which parts are built with AI and what that changes. An honest contractor will answer specifically, because they know. One who hides AI, or who only waves it around as a slogan, is telling you something important either way.
4. Compare scope, not amounts. A "website" in two offers is often two different products - I covered how different in freelancer or agency.
5. Check what is yours when it ends. Domain, hosting, code, content. If anything on that list stays with the contractor, you are not buying a website, you are renting one.

What that process looks like step by step on my side, from the first conversation to launch, is on the how we work together page.

Summary

AI genuinely lowered the cost of producing code - by roughly a third in my case - and it genuinely does build websites for free, as long as the website only has to exist. The price of a good website, meanwhile, tracks what AI does not do: decisions, copy and responsibility, and an honest contractor passes the saving from tools into quality rather than hiding it in margin. A free website is a legitimate, sensible option - provided you know you are taking on a website nobody is answerable for.

If you have a quote in front of you and are not sure what to make of it, book a free consultation - thirty minutes, bring that quote, and I will tell you what is in it and what is missing. Even if you end up building it somewhere else.