The slow part of building a website is usually not the building. It is working out what to build.
The usual sequence: a client asks for “something better looking”, it gets built, and seeing it real produces “this should stand out more” and “actually the customers we want are those ones”. Three rounds of revision and a fortnight has gone.
This is not clients being unprofessional. People find it genuinely hard to describe something that does not exist yet. You have to see it before you know whether it is right.
AI is undervalued at this stage. It is poor at deciding for you and exceptional at making you say what you mean, by asking. It also never tires of asking, and never pretends to understand a vague answer.
The list below can be pasted straight into ChatGPT or Claude. Working through it is worthwhile even if someone else builds the site — arriving with answers saves weeks.
Set one: establish what your business actually is
Open with this:
I’m building a company website. Don’t give me advice yet — interview me instead. Ask one question at a time until you can describe my business clearly. I run a (your industry) business in New Zealand.
Make it ask before it advises. The point is to get what is in your head out of it, not to hear general principles.
You will be asked, and do need answers to: who your customers specifically are (not “everyone”), what situation they are usually in when they need you, how they usually find you, how you genuinely differ from competitors, and which part of the business makes the most money.
That last one matters and is frequently not the same as the part you most want to show. Plenty of websites give their best space to whatever the owner is proudest of rather than whatever pays.
Set two: work out what visitors came to do
Based on that, list the types of people who would visit this site. For each, what do they most want to know, what are they worried about, and how many steps should it take them to contact us?
This surfaces things people had not considered.
A dental practice, for instance: one visitor is unhurriedly choosing a family dentist and will read about qualifications, the practice and reviews. Another is in pain and wants to know whether they can be seen today, where, and the phone number. Those needs conflict, and the home page has to hold both.
Or property management, where owners and tenants share almost no interests, so navigation should separate them at the door.
You know these things. Nobody has asked you.
Set three: settle the structure
Given all that, propose a page structure: which pages exist, which visitor’s question each one answers, and the order of sections on the home page. Use as few pages as possible.
Include “as few pages as possible” or AI will propose a dozen. More pages are harder to maintain and each needs someone to write it.
Then go through the result asking: which of these will I not have the energy to update within three months? That page probably should not exist. A page nobody updates is worse than no page.
Set four: draft the content
Write the home page copy. Rules: no filler like “committed to excellence”, “one-stop”, “professional team”; every sentence carries specific information; use numbers wherever a number exists. Afterwards, list which numbers I need to confirm.
That last instruction matters. AI will invent figures — it does not know how many years you have been trading or how many clients you have served, but it will produce a plausible-looking number. Have it flag them and check each one. Skip this and your website will claim something you did not know it claimed.
Set five: find what you missed
Act as a potential customer comparing three companies. Having read this structure and copy, what questions are still unanswered? What would make you hesitate?
This tends to produce the most valuable material. Common gaps: any indication of price, the service area, how soon work can start, what happens if something goes wrong, whether there is a contract.
Then one more:
If I could keep only three sections and delete everything else, which three, and why?
That forces priority. You will find that something you insisted on does not make the top three.
What you end up holding
After those five sets: a clear description of the business, the visitor types and their needs, a page structure, a first draft of the home page, and a list of figures to verify.
That is a brief. Taking it to a web company changes the conversation entirely — you discuss decisions instead of explaining what you do. Quotes get more accurate too, because the other side can see where the work is.
If you intend to build it yourself, it is equally the right starting point. Generating pages from this produces something very different from generating them from “make me a nice website”.
Two cautions
Discount its industry conventions. When it says “restaurant sites usually need online booking”, that “usually” is summarised from a great deal of material and may not fit you — perhaps you take no bookings and everyone walks in. Any advice beginning “usually” or “generally” needs checking against your actual situation.
It does not know local conditions. Market size in New Zealand, how the Chinese community here behaves, what local competitors are actually like — its grasp is thin and sometimes simply wrong. Do not take local judgements on trust.
We run a version of this before taking a project on, at no charge. If you have already done it yourself, that stage goes much faster and we start from the parts you have settled.