Most people only hire a web designer once every few years, so it is not something you get much practice at. You are spending real money on something you cannot fully judge until it is finished, and half the industry speaks in jargon that makes it hard to compare one quote against another. No wonder the whole thing feels like a bit of a gamble.
I have picked up a fair few clients over the years who came to me after a first website went wrong. The pattern is nearly always the same. The build itself was fine. What caught them out was everything around it: who owned the site, what the ongoing costs were, how quickly they could get a change made, whether it was ever going to bring in any work. Those are the things that decide whether you are happy in two years, and they are exactly the things that rarely come up in the sales chat.
So this is less a list of technical boxes to tick and more the conversation I would want a friend to have before they signed anything.
Start with what you actually need the site to do
Before you speak to anyone, get clear on the job. A website is not one thing. A joiner who wants the phone to ring needs something very different from a shop selling online, and both are different again from a consultant who mainly needs to look credible when someone Googles their name.
If your goal is enquiries, say so plainly at the first meeting and watch how the designer responds. The good ones will start asking about your customers, how people usually find you, and what a decent month of work looks like. The ones to be wary of will steer straight to how it will look, or start listing features. A pretty site that nobody contacts is a failure dressed up as a success. This is really the heart of it, and it is why we build every site around turning visitors into enquiries rather than just looking the part.
Should you just use an AI website builder?
It is a fair question now. You can open an AI website builder, type a few prompts and have something on screen before lunch. The colours will be fine, the sections will look familiar, and the copy will sound vaguely professional.
AI is useful, and any agency pretending otherwise is not paying attention. It can help with early ideas, page outlines, content drafts, image editing and repetitive technical checks. But it has not removed the need for judgement. It does not know your margins, which enquiries are worth having and which waste your time, why a customer in Cheltenham compares suppliers differently from one in Bristol or Stroud, or what proof your buyers need before they trust you. Deciding what the site needs to achieve, structuring pages around what people search for, writing copy that answers real objections, and making sure the analytics tell the truth is the real work behind a good website. AI can support that work. It cannot own it.
The same goes for templates. There is nothing wrong with them, and a small business does not always need a fully bespoke design. The problem is when the template decides the strategy: big hero section, three icon cards, vague headline, stock photo, contact form at the bottom. It looks like a website, but it does not explain why someone should choose you.
Look at their work, then look a little harder
Everyone will show you a portfolio, and every portfolio looks good, because designers only put their best three or four jobs in it. That tells you they can make something attractive. It does not tell you much else.
Push past the screenshots. Ask for a couple of sites they built a year or two ago and go and visit them on your phone. Are they still up? Do they load quickly? Does the business still seem to be trading? A designer whose older work has rotted or been replaced is telling you something. And if you can, ask for the name of a client you could actually ring. A two-minute chat with someone who has lived with the finished site for a year is worth more than any case study.
It is also worth checking they have built for businesses roughly like yours. Someone who mostly does restaurants and photographers may not understand how a trade customer searches, or where things like Gas Safe or NICEIC registration need to sit to build trust. It is not a dealbreaker, but it is a fair question.
Ask how they handle speed and accessibility
These are the parts you cannot see in a screenshot, and they are where cheap websites often fall down.
Speed is not a technical detail. Large images, bloated scripts and heavy animation all slow a site down, and on a phone with patchy signal that can be the difference between an enquiry and a bounce. Google's Core Web Vitals measure real user experience, and one of them, Interaction to Next Paint (INP), measures how quickly a page responds to clicks, taps and keyboard input. Google's own guidance says an INP of 200 milliseconds or less is good, while anything above 500 milliseconds is poor. A website is not a poster. People tap menus, submit forms and press phone numbers, and if it feels sluggish, trust drops before you have explained anything. A good designer thinks about this before launch, not after.
Accessibility is part of building a website properly, not a nice extra. The W3C published WCAG 2.2 as a recommendation in October 2023, and it covers things like colour contrast, keyboard navigation, focus states, form labels, error messages and readable text. A site can look acceptable in a screenshot and still fail when someone uses the keyboard, increases the text size, uses a screen reader or fills in a form on a phone. It overlaps with good business sense, too: clear labels, better contrast and forms that explain their errors help everyone, and a site that is easier to use usually converts better.