Solving the hard part: what a website actually needs in 2026

The gap between having a website and having one that works is the uncomfortable point this whole series rests on. You can build a website in an afternoon, but that was never the part that decided anything. Beneath that sits something more basic still: why you need a website in the first place. Whatever brings a person to you - a search, a friend, a post, an answer from an AI - the moment they decide whether to trust you happens somewhere you own, and nothing else quite finishes that job.

Both are argued in full earlier in the series; this one takes them as its starting point. That leaves the question you came for: what does the part that matters look like? What does a website need to do its job?

The obvious thing to do would be to give a checklist, that's what most articles that try to answer this question do - ten things every website must have, fast and mobile-friendly and a clear call to action, tick, tick, tick. But you can satisfy every line on a checklist and still be left with a website that does nothing. The dud that ticks every box is practically a genre. A website isn't a pile of features; it's one thing with one job, and the only way to know whether it works is to watch it do the job.

So we won't make a list. We'll follow someone in.

What follows is the bar a website has to clear in 2026, whether you built it yourself one wet weekend or paid a professional to. Some of it has been true since the web began. Some of it is true only this year, and will want revisiting next. I'll say which is which as we go.

The job

Orient, trust, act. In that order, because there isn't another one that makes sense.

Strip everything else away and a website has one job. It takes a person who has never heard of you and, if all goes well, turns them into someone who gets in touch. It does that the only way anyone is ever persuaded of anything: it helps them work out what you are and whether you're for them, then gives them reason to trust you, then shows them what to do next. Orient, trust, act. In that order, because there isn't another one that makes sense.

Everything a good website needs is just something that job demands. So instead of counting the parts, let's watch one work. Here is a person arriving, and here is what has to be true at each step for her to still be there at the end.

Follow her in

Let's take the same woman from the first piece in this series: the one who got your name from a friend at the school gate and reached for her phone. This time, let's say the site she lands on is a good one, and watch what that means.

A good site has thought about her in advance, and the reassurance is waiting where she reaches for it.

First she has to land on it at all. She doesn't type your address, because nobody does that any more. She searches for the thing she needs in the town she needs it in, and your site is there to be found, because it was built to be found. How that happens we'll come to in a moment, because she never sees it. For now, she's arrived.

The first few seconds go on a single question, though she'd never phrase it this way: am I in the right place? A good site has answered it before she has to ask. At a glance she can tell what you do, who you do it for, and that this is a real business rather than a template wearing your name for the afternoon. Get this wrong and there's no second go at it. She's back to the search results, and as far as you're concerned she never existed.

If she stays, she starts doing the thing you never get to watch. She looks for reasons to trust you, and reasons not to. This is the stretch that does the real work, and it's the one cheap sites skimp on hardest, because it's the most work. She wants the answer to the specific worry in her head, and her worry isn't the same as the next visitor's. Do you actually do this exact job? Have you done it for someone like her, in a house like hers? What will it be like to deal with you? Is there a real person here, with a name and a face, or only a form and a stock photo of a handshake? A good site has thought about her in advance, and the reassurance is waiting where she reaches for it. A bad one makes her hunt, and hunting is just a slower way of leaving.

Only when that's settled does the easy part arrive. She decides to act, and the site makes acting obvious. The next step is right in front of her, it's the step she wants, and it asks no more of her than it needs to. The number of sites that earn all the trust and then hide the way to get in touch is a small tragedy of its own.

That's the whole encounter. A minute, maybe two. If every step held, she's an enquiry in your inbox. If one step failed, she's gone, and you'll never find out which one.

The part she never sees

Now go back through that minute and notice everything she didn't.

She didn't wait. The page was there almost before she'd finished tapping, because a fast site is built to be fast and a slow one quietly loses the people who will never tell you they left. She didn't pinch or zoom, because it worked on her phone as though the phone had been the point all along, which today it has to be. She found you in the first place because the site was built so that a machine could understand what it was about, which is the unglamorous plumbing underneath the thing everyone calls SEO. And that same plumbing increasingly decides whether an AI will put you forward when someone asks it, because an AI can only repeat what it has been able to read and trust about you. None of this showed. All of it was holding the floor up.

The easiest corner to cut, and the hardest to be caught cutting.

There's more she didn't see. Had she been using a screen reader, the site would have worked for her too, which is partly basic decency, partly the law, and partly the very structure that machines reward anyway. And the photographs were real: your work, your face, your van. In a year when the web is filling up with generated images, people have started, half-consciously, to distrust the ones that look too perfect. Being real is quietly becoming an advantage, for no better reason than that so much no longer is.

This is the part worth slowing down on, because it's where the bad website hides. Everything in this section is invisible to the person it's working on. You cannot tell, glancing at a site, whether it's quick enough, built properly underneath, legible to a machine, honest in its pictures. Which makes it the easiest corner to cut and the hardest to be caught cutting. A site can look perfectly fine and be failing every one of these, and the only sign will be the enquiries that never arrive. That is how a website somebody was paid good money to build can still do nothing at all. Not because it looked wrong. Because the work that was skipped was the work nobody can see.

One caveat about this section especially: it dates. The principle, that the invisible work is what carries the visible, is permanent. The specifics move. What counts as fast, what it takes to be understood by a search engine or quoted by an AI, where the bar sits for being readable to everyone, all of it shifts, and shifts faster than anything else here. That's why this piece keeps a year in its title. Read it a few years from now and trust the principle over the particulars.

It doesn't stay done

A website isn't a thing you build. It's a thing you run.

Suppose you have all of that. The site does its job, the visible part and the hidden part, on the day it goes live. You're still not finished, because a website isn't a thing you build. It's a thing you run.

Launch day is the best it will ever be if you then leave it alone. The world it describes keeps moving. Your prices, your services, the team photo that's a face out of date, the competitor who did all of this six months after you and is now closing the gap. Search engines change the rules without telling you. Things quietly break. A website is less an ornament you hang on the wall and admire than an employee you'd never dream of leaving alone for a year, with no instruction and no idea whether they were still turning up. It needs someone whose job is to keep it true, keep it working, and keep it ahead. Skip that, and even a site that was excellent on launch day drifts, over a year or two of neglect, back into being one of the dead ones we started with.

So, what does a website actually need?

None of this is a wish-list, and none of it is the expensive end of the market. It's the floor. It's the line beneath which a website isn't worth building, because below it the thing simply won't do its job, and a website that doesn't do its job is worse than not having one, since it cost you money to be ignored. Everything here is what "worth building" means. Above the floor there's all the room in the world for ambition. Beneath it there's no point starting.

So, what does a website actually need? It needs to do one job well: to carry a stranger from never having heard of you to getting in touch, in the order people have always been persuaded, with the visible work done where they can feel it and the hidden work done where they can't, and then kept alive long enough to go on mattering. That's the whole bar.

Which leaves you an honest choice, and either answer is a good one. If you've got the time, and you find you really do like this sort of thing, you can clear that bar yourself, now that you know where it sits. And if you'd sooner spend those hours on the business you're already good at, you find someone to clear it for you, knowing now what to ask for and how to tell whether you've got it. The one thing you can't really do, having read this far, is what most people do: build the easy part, skip the rest, and wonder months later why the website never brought them anything.

People think websites don't matter. They do. You just needed to know what makes one matter.