If you have been on X this week, two comments keep colliding. One person compared vibe coding to 3D printing: you can produce a convincing object fast, but a prototype is not a manufactured product. Another pushed back on the "you can vibe-code it in five hours" claim by asking what five hours, five weeks, and five months actually get you.
That argument is usually about apps. It is just as useful for marketing sites.
A coding agent or an AI website builder can give you a polished page in an afternoon: hero, social proof, a pricing table, a form. In the browser it looks like a company. That is a real achievement. It is also, in most cases, a landing page. A landing page and a website are different jobs, and mixing them up is how a five-hour win turns into a messy six-month rewrite.
A landing page has one job
A landing page is one URL, built to get one action from one kind of visitor. Navigation is thin or missing on purpose, so nothing competes with the signup, the booked call, or the pre-order. Traffic usually comes from somewhere you already control: an ad, an email, a post, a link you share.
A website is a set of pages under one domain, linked by a menu, that has to serve more than one visitor and more than one question. The homepage is the front door of that set. It can be designed to push one action, but it is not the same thing as a campaign page.
If that distinction already feels familiar, skip ahead. If it does not, it is the landing page vs website decision most vibe-coded launches skip.
When five hours is enough
Ship a single page when all of these are true:
- You have one offer and one audience.
- You need one result: a waitlist, a demo request, a sale.
- Most visitors will arrive from ads, email, or links you send them.
- You can keep one page honest — prices, claims, and the form destination — without a CMS debate.
That is a good first ship. It is how you test demand without inventing a seven-page information architecture you will throw away.
Two cautions, even then.
Put it on a URL you plan to keep. A rented preview subdomain is a demo, not a launch. And decide whether the page should be indexed. A campaign page that repeats your main offer often belongs behind a noindex robots tag (do not block it in robots.txt; Google has to crawl the page to see the tag, and ad checkers have to reach it too). A page you want found for your brand or core offer should be indexable, and it will only compete well for one main search need.
When five hours is not a website
Start with a small site, not a stretched landing page, when any of these show up:
- You sell more than one product or service, or you serve more than one audience.
- People need to find you through search, not only through links you share.
- Buyers ask about process, pricing, the team, or proof before they act.
- You already have a second article, a case study, or a comparison worth publishing.
A single page can rank for one main query and its close variations. It cannot do a good job of several distinct searches. One URL trying to cover pricing, three services, and a blog-length FAQ becomes hard to update and easy for a crawler to misunderstand.
The other failure mode is the opposite: asking the model for a 12-page site in one sitting. You get thin service pages, duplicated intros, and a menu that looks complete while every URL says the same thing. Five strong pages beat fifty interchangeable ones, especially on a new domain.
A practical split: five hours, five days, five weeks
Treat the work as stages, not as one prompt.
Five hours. A landing page with a real offer, a real next step, and copy you would actually send to a customer. No fake metrics. No invented testimonials.
Five days. The unglamorous production bar: a domain you control, HTML that still contains the copy when you curl it, a working form, and one person who can edit a typo. If search matters, add a title, a unique description, and a sitemap entry. If you already know you need a small site, this is when you lock the URL pattern so you are not redirecting next month.
Five weeks. The first set of pages that each own one job: the homepage (or the landing page you kept at the root), one page per core offer, prices, about, and contact. That set is the backbone of a zero-to-indexed launch plan. Add a menu only when there is somewhere worth linking.
You do not have to wait five weeks if the answers are already in your head. You do have to resist generating the shell of a large site before those pages have something true to say.
Grow the page. Don't replace it.
The low-risk path is boring, which is why it works.
- Launch the page you need now on the URL you intend to keep. If this page will become the homepage, put it at the root, not on
/lp-v3. - When a second offer, a second audience, or a question the page cannot answer shows up, add a new URL for that job. Link it from a simple menu.
- Keep the original URL in place. Redirects are for addresses that actually change, not for a redesign you did not need.
- Give each new page a self-referencing canonical and add it to the sitemap.
- Stop when the next page would only exist to look busy.
If you later run a paid campaign, make a separate landing page for the ad. Do not turn the homepage into that campaign, and do not leave a temporary offer sitting on the root when the campaign ends.
An AI website builder is useful here when it helps you add those pages on a domain you own, with copy that starts from your actual offer rather than a generic template. It is a poor fit when the prompt is "build me a complete corporate site" and nobody checks whether each URL has a reason to exist.
The test before you call it done
Ask three questions:
- What is the one action this URL is for?
- If a stranger arrived from search, which question does this page answer better than a generic alternative?
- When the next page is needed, can I add it without moving this one?
If you can answer those, the five-hour page was worth shipping. If you cannot, you do not need a bigger prompt. You need a clearer job for the page you already have.
