The short answer
Send the live URL, not your logins. Ask for a fixed price in writing before anything starts, and ask to see a homepage design before you commit — that single question filters out most of the people you’d regret hiring. Expect a rebuild rather than a patch, because untangling generated code usually costs more than building it cleanly. And get two or three quotes, treating a dramatically low one as a warning rather than a win.
You’re in a normal place, not a failure state
Worth saying first, because a lot of people arrive here slightly embarrassed: building something with an AI tool and then hitting a wall isn’t a mistake. It’s the tool working as designed. You now know what you actually need, which you didn’t know before, and you know it without having spent months and a large budget finding out. That’s a genuinely good position.
What’s changed is the question. It’s no longer “can I get something online” — you did that. It’s “will this hold up now that customers are using it”, and that’s a different job with different requirements.
Step one: decide what you’re actually buying
These are three different jobs, and they attract different people at different prices. Knowing which one you want before you start asking around will save you a lot of confusing quotes.
A rescue. The site basically works, but something specific is broken — the contact form drops enquiries, it’s painfully slow, it falls apart on a phone. Narrow, well-defined, cheapest. Right when the underlying thing is sound.
A rebuild. Same site, same content, built properly from scratch. What most people need once real customers are involved, and often cheaper than a rescue despite sounding bigger, for reasons covered below.
A redesign. You’ve realised the site is wrong, not just badly built — wrong structure, wrong message, wrong idea of who it’s talking to. The most expensive, and the only one that starts with a conversation about your business rather than your code.
If you’re unsure, say so when you ask. “I don’t know if this needs fixing or replacing” is a perfectly good brief, and how someone answers it tells you a lot about them.
Why “just fix what’s there” often costs more
This is the most counter-intuitive part, and it’s worth understanding so you can tell a real recommendation from an upsell.
Reading unfamiliar code is slower than writing new code. That’s true of all code, and it’s especially true of code generated in a long chain of prompts, because there’s no design intent behind it to reconstruct — no reason why things are arranged as they are, no comments explaining the tricky parts, sometimes several half-finished approaches to the same problem sitting side by side because the model changed its mind mid-conversation.
So a developer looking at your generated site is often choosing between a week of archaeology and three days of building. The three days is usually both cheaper for you and better for you.
The honest exception: if the site is largely fine and the problem is narrow, a rebuild is overkill and you should push back. Ask for both options priced separately. Anyone who won’t explain their reasoning in plain language is either not confident in it or not being straight with you.
What to send
You need less than you think. This is enough for a real quote:
- The live URL. The single most useful thing. A developer can tell an enormous amount from the front end.
- One sentence on what the site is for. “I run a remedial massage clinic in Sydney and I need people to book online.”
- The three things it absolutely must do. Not twenty. Three.
- Whatever’s currently broken or embarrassing you. Be blunt. Nobody’s judging the AI build.
- Any hard deadline, and what’s driving it.
What you should not need to send: your platform password, account access, or payment details. A quote comes from looking at the site, not from getting inside it. If someone wants access before they’ll give you a price, that’s worth pausing over — it isn’t automatically dishonest, but there’s rarely a good reason for it.
The six questions
Ask these of everyone you’re considering. They’re chosen because a good developer answers them easily and a bad one gets vague.
1. “Can I see a homepage design before I commit to anything?”
The best single filter there is. Someone willing to show you the direction before taking your money is confident in their work and comfortable being judged on it. Someone who wants payment before you see anything is asking you to take all the risk. Ask this first — it’s also the question that tells you the most per minute.
2. “Is that a fixed price, or an estimate?”
Fixed means the overrun risk sits with them. Hourly with no ceiling means it sits with you, and web projects overrun. Get the number in writing before work starts.
3. “What happens when something breaks in six months?”
You want a person’s name and a realistic response time. “Submit a ticket” from a one-person operation means there is no answer.
4. “Who will hold the domain and the hosting?”
There’s no single correct answer here — plenty of good arrangements exist. There is a wrong answer, which is nobody being clear about it. Whatever the setup, you should know how to get your site somewhere else if the relationship ends.
5. “Can you explain what happens when someone submits my contact form?”
Deceptively powerful. You’re not testing whether you understand the answer — you’re testing whether they do. Where does the data go, who can read it, what stops a bot submitting it ten thousand times. Anyone who can’t explain this in plain language is prompting an AI and hoping, which is exactly the situation you’re trying to get out of.
6. “What would you tell me not to build?”
Anyone who says yes to every idea hasn’t thought about your business. A good answer to this is worth more than a good portfolio.
Red flags
- Wanting your logins before quoting. Covered above.
- No design until you’ve paid. See question one.
- Hourly with no ceiling. You’re absorbing all the risk of their estimate being wrong.
- No verifiable client work. Named businesses with links to live sites you can open. Screenshots in a PDF prove nothing — anyone can screenshot anything.
- A quote dramatically below the others. Usually it’s buying you a first draft rather than a finished piece of work. The Airtasker version of this is worth reading — the price difference is real, and so is what it’s paying for.
- Vagueness about what’s included. “SEO included” means nothing on its own. Ask what specifically.
- Pressure. Countdown offers and “two slots left” on a custom build are theatre.
Freelancer, agency, or marketplace?
A freelancer or one-person studio gets you the person doing the work, directly. Cheaper, faster decisions, no account-manager layer. The genuine risk is the bus factor: one person, one holiday, one illness.
An agency gets you continuity and cover, at a higher price, with the person you’re briefing usually not the person building. Right when the project is large or you need contractual weight behind it.
A task marketplace gets you the lowest price, and the trade-off is real: the hours needed for discovery, testing and follow-through don’t fit inside a very low bid, regardless of how skilled the individual is. Fine for something small and well-defined. Risky for the site your enquiries depend on.
A reasonable process to expect
However you go, the shape should be roughly:
Discovery — a conversation about the business, not just the site. Design and quote — you see the direction, then agree scope and a fixed price in writing. Build — with you seeing progress, not disappearing for a month. Launch and handover — you know how to reach someone afterwards.
If a proposal skips design-before-commitment or won’t fix the price, ask why. There may be a good reason. There may not.
This site is one place that does this work — rebuilding AI-generated and drag-and-drop sites properly, for Australian service businesses. But the questions above are worth asking whoever you hire, including if that’s not us. Send a URL and you’ll get a straight answer on whether a rebuild is worth it — including when it isn’t.