The 12-Month Bill Nobody Shows You: What a 'Free' Site Actually Costs vs. One Claude Builds You
Michael Sacca•Every pricing page is a trailer for a movie you haven't agreed to watch yet. The number in the big font is what you pay on day one. The number that actually matters is what lands on your statement in month twelve — and that number is almost never on the pricing page. If you're comparing a free website builder against a site Claude builds you, the comparison only becomes honest when you model the whole year.
We've already taken apart what "free" quietly doesn't cover in Free Website Making Websites: What 'Free' Actually Buys You (and What It Quietly Doesn't), and what those builders look like after the honeymoon in Day 30 Is the Real Test: What Free Website Builders Look Like After the Launch Honeymoon Ends. This post is the spreadsheet version: actual line items, month by month, so you can see where the curves cross.
The signup price is the teaser rate
The pattern across the major builders is consistent, and it's worth naming before we put numbers anywhere. A free tier gets you a subdomain, builder branding, limited pages, and often no custom domain. The moment you want a real site — your own domain, no ads, e-commerce, or in many cases any meaningful storage — you're on a paid tier. That tier is priced attractively for the first billing cycle.
Then renewal happens. Introductory discounts typically expire at your first renewal, and the renewal price is frequently a meaningful step up from the teaser. The exact figures vary by plan and by when you catch a promotion, so I won't pretend there's one universal number — but the shape is universal: cheap year one, noticeably more from year two, billed annually, and the cheapest tiers usually require you to commit to 12 months up front to get the good rate.
Meanwhile, the costs that get zero marketing at all: the domain itself is sometimes "free for year one" and renews at full registry price plus a markup afterward. Email inboxes are usually an add-on. Remove-the-badge used to be a premium feature on some platforms. None of this is a scam. It's just a pricing structure that only reveals itself over time — which is why nobody shows you the 12-month bill.
The other side of the ledger: what an LLM-built site costs
An AI-built site, in the workflow this blog keeps coming back to, is a folder of static files. That changes the cost structure entirely. Here are the line items:
1. The LLM itself. If you're already paying for Claude Pro or ChatGPT Plus — both run around $20/month at list price, though check current pricing — the marginal cost of building your site is close to zero. You'd have the subscription anyway. If you're using API credits for agentic builds, spend varies wildly with how much you iterate; a modest landing page might cost a few dollars in tokens, a multi-page site with heavy revision loops more. This is the one line item that genuinely varies, so treat any single figure you see — including mine — as an estimate, not a quote.
2. Hosting. Static hosting on Netlify, Cloudflare Pages, Vercel, or GitHub Pages is free or near-free for a personal site's traffic levels. You hit paid tiers at volumes a solopreneur rarely reaches in year one.
3. The domain. Same as everyone else — registry price, roughly ten to fifteen dollars a year for a .com depending on the registrar. No builder markup, no "free first year" bait, and full portability, which we covered in The 6 Alternative Domain Hosts That Actually Play Nice With AI-Built Sites.
4. Your time. This one cuts both ways. A builder is faster if the template fits. Claude is faster the moment you want something the templates don't do — and, crucially, every future change is a prompt instead of a plan upgrade.
Run the year
Let's model a modest scenario: a small business landing page plus blog, custom domain, no e-commerce. I'll use ranges and you should substitute the current numbers from whichever builder you're eyeing — the method matters more than my placeholders.
Builder path, 12 months:
- Paid tier after outgrowing the free wall: one year at the promotional annual rate (often presented as a monthly-equivalent figure, billed upfront)
- Domain: free year one, renews at full price in month twelve
- Email: add-on, monthly
- Cost to change anything structural that your plan doesn't include: another tier bump
Claude path, 12 months:
- LLM subscription: ~$20/month, but presumably a sunk cost if you already use it
- Static hosting: ~$0 at realistic year-one traffic
- Domain: full registry price from day one, no teaser
- Iteration cost: your prompts
Add it up and a first year can come out close to a wash — the builder's teaser pricing is genuinely good, and I'll say so. The break-even usually isn't in year one. It's in year two, when the builder renews at the full rate and every plan-limit you brush against becomes an upsell, while the AI-built site's costs stay flat: domain, hosting, and a subscription you were probably paying anyway.
The costs that don't appear on any bill
There are three hidden line items that price comparison sites never model, and they're the ones that actually decide this.
Switching cost. Builder exports are famously imperfect — some give you clean static files, many give you something that only re-imports into themselves. The "free" site that's expensive to leave isn't free; it's financed. An AI-built site in your own repo has a switching cost of approximately nothing, because your AI developer doesn't need your codebase to rebuild it — but you'll be glad you own it.
Change cost. With a builder, every "can we add a section to the homepage" is your evening or a designer's invoice. With Claude, it's a prompt and a deploy. Over twelve months of real business — a new service page, a seasonal banner, a pricing change — this is the biggest number on the spreadsheet and the hardest to predict in advance. The workflow, not just the first post is what makes it cheap, month after month.
Failure cost. When a builder has an outage or deprecates a feature you depended on, you wait. When your own static site breaks, you and your LLM fix it in the same session — which is also why we push people toward letting the LLM run the whole deployment pipeline instead of stopping at generated code.
A template for your own 12-month model
Before you commit to either path, fill this in with live prices. If a cell is "unknown," that's your answer about that vendor.
LINE ITEM BUILDER PATH CLAUDE PATH
Tier (year 1, promo) $____ n/a
Tier (renewal rate) $____ n/a
Domain (year 1 / yr 2) $____ / $____ $____ / $____
Email hosting $____ $____
Hosting (year 1) included? ~$0 (static)
LLM subscription n/a $____ (or $0, sunk cost)
Plan bumps you predict $____ $0
Time per future change ____ hrs 1 prompt
Export path ____ repo you own
Two honesty notes. First, if you'd never pay for an LLM subscription otherwise, the builder's teaser year can genuinely win — Claude's advantage assumes the subscription is already part of how you work. Second, if you need e-commerce, booking, or accounts, neither of these is the right comparison; that's an application, not a static site, and the calculus changes.
But for the most common case — a real site with your own domain that you'll actually keep updating — the pattern holds. The builder shows you a cheap first month. The repo shows you a cheap forever, minus the subscription you already have. Nobody shows you the 12-month bill because the people selling the teaser rate would rather you didn't do this math.
Do the math anyway. It takes ten minutes and it's the highest-leverage spreadsheet you'll make this year.