Why we do not build business websites on Lovable, Bolt or v0
These platforms wrap an AI model in hosting and a subscription. They are fast, and for a throwaway prototype the speed is the whole point. For a website a business depends on they are a layer that adds a dependency, removes control, and is designed to skip the only step that makes a site worth building.
The mechanism
Prompt to live, with nowhere for a decision to happen.
This is an argument about how the tools are designed, not about the people who built them. The path these products optimise for is description in, deployed website out, as fast as possible. Every step that makes a website distinctive has been engineered out of that path, because every step is friction.
- 01
There is no strategy step
Nothing in the flow asks what this business can claim that its competitors cannot, or what one action the page exists to produce. Without that, the model has no choice but to return the middle of everything it has read.
- 02
There is no review step
The generated code goes live because that is the product. Nobody reads it, so nobody catches the form that does not deliver, the heading structure that means nothing, or the contrast that fails.
- 03
There is no evidence step
The site launches with nothing measuring it. The weeks when behaviour is easiest to read and the fixes are cheapest pass with nobody watching, and you cannot get them back.
- 04
So the output converges
Same models, same training data, same absent steps. Ask three of these tools for a site in your industry and you will get three versions of the same page. Your competitor can produce yours this afternoon.
Redundant
The platform adds a dependency and takes away the ceiling.
Here is the part that decides it for us. The platform is not giving you access to some better model. It is giving you the same models everyone else can reach, wrapped in hosting you cannot leave easily.
- The intelligence is not the platform. Cursor, Claude Code and Codex reach the same models and build into a repository you own outright.
- So the layer contributes hosting and a subscription, and charges for both, for as long as the site exists.
- In exchange your website's future is tied to a company's pricing and roadmap, which is a decision about your business you did not get to make.
- There is a ceiling. When you need something the platform does not do, the answer is not "build it", it is "you cannot".
- Verify the export before you believe in it. Exporting files nobody else can run is not portability, and the time to find out is now, not when you want to leave.
What we use instead
The same AI, pointed at code you own.
We are not arguing against AI writing websites. We use it on every build. We are arguing about where the result lives and what happens between the prompt and the launch.
- Professional tools on a real codebase: Cursor, Claude Code and Codex, in a framework any developer can pick up.
- A half day of strategy first, deciding what this business can say that nobody else can, because that is the part no model can supply.
- Every build read by a person before it ships, and measured against real checks for speed, accessibility, search and form delivery.
- The repository, the domain and the hosting in the client's name from day one, with no subscription to us for the site itself.
- Five years of watching which sites earned and which merely existed, which is what tells us where the generated draft is wrong.
FAQ
Questions people ask about this
Should I build my business website on Lovable or Bolt?
Not one your business depends on. The site would sit on someone else's platform, subject to their pricing and roadmap, and the path those products are designed for runs from prompt to live with no step where anyone decides what makes your business the obvious choice. You end up with a competent page that says what every page in your industry says. If you want AI to build your site, and it should, have it build into a codebase you own.
What is the difference between Lovable, Bolt, v0 and Replit?
Less than the marketing suggests, because they reach the same underlying models. Broadly, Lovable and Bolt aim at complete apps deployed immediately, v0 comes from Vercel and leans toward interface components that slot into a Next.js codebase, and Replit is a development environment with an agent attached. For a business website the distinction that matters is not which generates the nicest first draft, it is whether the result ends up somewhere you control.
Can you export the code from an AI website builder?
Some form of export usually exists, and it is the single most important thing to verify before you commit. Do it on day one: export the project and get it running somewhere else entirely. An export that produces files nobody else can run is not portability. What is offered varies by platform and changes over time, so test it rather than trusting a feature list.
Why would I pay an agency when AI can build a website?
Because AI builds ours too, and the building was never the expensive part. What you are paying for is five years of launching websites for other businesses and seeing which ones earned their keep, and a strategy session that decides what you can claim that your competitors cannot. A model can only return the average of what already exists, and an average cannot be the obvious choice. That is the gap, and no amount of prompting closes it.
Are AI website builders worth it for a small business?
For a genuine throwaway, a prototype you will discard, or an internal tool only your team touches, the speed is real and the stakes are low. For the website that decides whether a stranger calls you, it is the wrong foundation: you get a page indistinguishable from your competitors', hosted somewhere you do not control, with nothing measuring whether it works. A small business has less margin for that, not more.
Start your brief
Start your brief
Tell us what you're launching and what it needs to do. We reply within one business day with a fixed price and the first available Monday.
- Websites from $15,000 ex GST, fixed. The exact figure comes with the reply.
- A reply within one business day, from Michael, not a form robot.
- A thirty-minute call to hear what the site has to do.
- A fixed-price scope, sent as a web page with a version number.
- Kick-off on the next available Monday, held for you. We take four studio builds a month.