What a Website Really Costs in the Age of AI
AI has made building a small website cheap. It has not made owning one cheap, and the pitfalls have moved from the build to what comes after.
I first wrote this in 2018, after too many conversations with small business owners who had been handed wildly different quotes for what sounded like the same website. Back then the cost was mostly the build. Today an AI tool can produce a decent first version of a small site in an afternoon, so the old price list no longer means much and I have dropped it. What has not changed is where people get caught out. It has just moved.
What Got Cheaper
The build itself. A one-pager, or a simple company site with a few pages, can now be drafted by an AI site builder or by a coding assistant like Claude from a short description. For a barbershop, a tradesperson or a local shop, that is genuinely good news. The page that used to cost days of someone’s time now costs a few hours of yours, and it is often good enough to launch.
The pricing ladder I used to describe, one-pager, company site, online shop, app, still holds as a ladder. Each step up adds work that is not about typing code: gathering content, connecting real systems, getting payments approved, testing and securing it. AI shrinks the typing. It does not shrink the rest, and the rest grows quickly the moment money or customer data is involved.
What Did Not Get Cheaper
- Knowing what the site is for. A clear offer, a clear next step for the visitor, words that sound like you. A model will happily write fluent filler. Deciding what to say is still your job.
- Real content. Your photos, your prices, your reviews, your story. Generic AI text and generic stock images make every small business site look the same, and visitors notice.
- Hosting and your domain. Still small, still recurring.
- Upkeep. Updates, security fixes, the occasional change. A site nobody looks after becomes a liability, however quickly it was built.
- Payments and shops. The provider still has to approve your business, and the legal and tax rules are still your problem.
The New Pitfalls
Where your site actually lives. Many AI builders host the site for you, and some make it awkward to leave. Before you build, find out whether you can export the full site and its content, and what you would pay to keep it live. If you cannot take it with you, you are renting, not owning.
Code nobody understands. If an assistant wrote it, someone still has to maintain it. Generated code can contain forgotten test keys, unused packages, and shortcuts that work until they do not. Whoever looks after the site should be able to read it, and you should know who that is.
Quietly collecting data. Chat widgets, analytics, embedded forms and AI assistants often send your visitors’ details to a third party. Under UK GDPR and the cookie rules, that is your responsibility as the business owner, not the tool’s. Ask where visitor data goes, whether it is used to train anything, and whether you need a consent banner and a privacy page. If you add an AI chatbot, say so plainly to your visitors.
Cheap quotes that are really a prompt. Some suppliers now deliver an unreviewed AI draft and charge build prices for it. That is not automatically bad, but you should know. Ask what was generated, what a person reviewed, and who fixes it when something breaks.
Know Who Owns the Keys
This advice is unchanged since 2018, and it matters more now that so many tools are involved. Two things should be in your name, not your developer’s or a builder’s:
- Your domain. Register it to your company, with the login held by you. If someone else holds it, write down how you get it back. I have seen a business lose a domain it used for a decade because the person who registered it moved on.
- Your data. Customer lists, orders, content and accounts are yours. Know where they are stored and how to take a copy.
Where sensitive information changes hands, a short NDA keeps everyone covered.
What I Would Do Now
Use AI to get a first version up quickly, then spend the saved money on the parts a model cannot do: a clear message, real content, a review of what it built, and a person who will look after it. Own your domain and your data from day one, check you can leave any platform you use, and pick someone you can talk to.
If you are weighing a quote, or a draft an AI gave you, and want a second pair of eyes on it, say hello.
Jimmy MG Lim writes about software engineering and cloud infrastructure. Runs a small consultancy between London and Southeast Asia; say hello.