AI and websites

AI published your website in an afternoon. Choosing where it lives was done without you

The site is live before the first meeting ends, and three contracts were signed along the way: where it runs, whose name is on the address, and what it costs to leave. None of them appeared as a decision.

The signup form made three decisions before anyone asked you

Publishing an AI-built website now takes a credit card and an afternoon, and the hosting decision hides inside that speed. When the preview looks right, one click puts the site online. That click also chooses which platform the site lives on, whose name goes on the domain, and what happens to both when the subscription stops. None of these appear in the preview, and all three outlive it.

A joinery that publishes on the builder’s free address, registers the domain through the same signup form to the e-mail of whoever held the card, and moves on with its season, is the normal case, not a cautionary tale. Nothing is broken. The problem is that the three decisions were made by whoever filled the form fastest, and the business will be the entity that lives with them.

What the bundled setup genuinely gives you

The bundled hosting is not a trap; it is a good default that nobody revisits. The certificate issues itself, backups run on a schedule, updates happen without anyone phoning a server room at midnight. For a brochure site in its first year, that is real value and zero operations work, and AI made it better: the site deploys on save, the assets sit on a global network, and nobody had to learn what any of that means.

The intro price is real too. A free domain for the first year, hosting for the price of a lunch. The catch is not in the offer. It is in the absence of a decision: the moment of signup is the only moment when leaving was still cheap to imagine, and almost nobody imagines it while the site is still loading for the first time.

Where renting quietly turns into lock-in

The subdomain is the smallest problem and the most visible one. The address on the business card, the e-mail signature and the invoice header reads like a platform advertisement, and customers type it wrong because they drop the middle part. Moving to your own domain later is possible, but it splits the site’s history in two, and every ranking, every saved link and every printed QR code has to be redirected or rewritten.

The contract is held by an individual, not by the company. The platform account sits in the personal inbox of whoever signed up, on a card that expires in three years. When that person changes jobs, the website, the domain renewal and the form inbox all stop being reachable by the business — not because the platform locked them out, but because nobody ever transferred what was never formally the company’s.

The exit terms exist; they are simply read by nobody during signup. Year-two pricing, page limits on the cheap plan, export scope, what happens to collected contacts if the subscription lapses. These are written down in the terms the signup form made scrollable. The price of discovering them later is measured in weekends.

What survives a move, and what does not

The domain moves if, and only if, it is registered in your name. A domain registered through the platform to a personal address can usually be transferred, but the process runs through that personal inbox and takes weeks. A domain registered from the start at a registrar the company controls moves in days, and nobody has to be persuaded to cooperate.

Content moves; structure and history move only partially. Texts and images can be exported from almost any platform. The layout, the integrations, the forms wired to the platform’s own inbox, the analytics property with two years of history — these do not come along, or they come along as data that nothing on the other side reads. A move is therefore never a copy. It is a rebuild with the content pre-written.

Form submissions and customer records are the real hostages. If every enquiry the business received lives in the platform’s inbox, then leaving means either exporting contacts by hand and losing the thread, or staying. Businesses do not stay loyal to hosting. They stay because leaving looked expensive on the day they checked.

The five checks that take ten minutes before you publish

Run them while the signup form is still open, not after the launch announcement. Who is the registrant of the domain, and is it a company address rather than the account of whoever signed up? Does the plan support a custom domain from day one, and what does it cost in the second year? What exactly does the export contain — files, content, page URLs, form definitions? Where do form submissions land, and can they be forwarded to the company’s own inbox or CRM? What is the price after the intro period, and where is it written?

If the answers exist on the pricing page rather than only in the terms of service, the platform expects businesses to leave eventually and has made that survivable. That is the platform worth publishing on — even if the first version of the site is rough, because a rough site on an exit-friendly platform is a cheaper problem than a polished site on a dead end.

The decision worth making on purpose

Hosting and domain feel like settings; they are contracts, and they outlive every version of the website. Keep both in the company’s name, on accounts two people can reach, with a renewal date somebody actually sees. The site itself can be rebuilt in a weekend when the content is yours. The address, and the history attached to it, cannot be rebuilt at all — which is exactly why the choice of where it lives deserves the ten minutes the signup form never asks for.


Related service: Website and web application development