Building a website for SEO. Here's what's actually making a difference in 2026.

I spend my working life building websites for other people, while slowly neglecting my own website, which is the most predictable thing a designer can say, right?

So I am rebuilding it. Not the look, the findability. And because the rules have changed more in the last eighteen months than in the five years before that, I thought I would write down what I am actually doing as I do it and report on what's been actually working so far.

Why one page is never enough

One-pagers are great. Really. If you need a home base or a business card on the web, a single well-built page will serve you for years. But for search, one page means one shot at being found.

I had a home page and a single services page holding everything I do. Web design, branding and graphic design all lived in the same place, competing for the same attention and the same keywords. Now each one has its own page, with its own title, its own focus keyword and its own meta description. Three pages, three entry points, three different searches I can actually show up for.

One caution, because this advice gets taken too far: it only works if each page is genuinely different. Three thin pages saying near-identical things are worse than one good one. Google has been unimpressed by volume for a long time.

Why I am adding a blog when I already write elsewhere

I already write a newsletter. It has readers, it has an archive, and it does precisely nothing for my website, because that traffic, those keywords, and that authority all live on somebody else's domain.

A blog on your own site is different. Articles that answer real questions are how people find you months before they are ready to hire anyone. Somebody searches "how much does a website cost". You want to be the one who answers that question.

The thing that has changed here is the standard. Generic posts that rehash whatever is already ranking get ignored now, and there are a lot of them, because they take four minutes to generate. What still works is the thing only you can write. Your real projects, your real numbers, the opinion you would actually defend in a meeting.

Why I am adding FAQs that Google stopped rewarding

This is the one that surprised me. In May 2026, Google stopped showing FAQ rich results. Those expandable question dropdowns under search listings are gone, Search Console dropped the reporting in June, and the API metrics went in August. For most of us who had added FAQ sections to chase that little bit of extra real estate in search, the reward quietly disappeared.

And I am adding more of them than ever. Because AI does not read a website the way a person does. It is looking for a clear question with a clean, self-contained answer directly underneath, something that still makes sense when it is lifted out of your page and dropped into somebody's answer. An FAQ is that shape already. You are handing it over pre-packaged.

One rule: do not repeat the same question on five pages. Each FAQ should belong to the page it is on and answer something specific to it. Five copies of "how much does a project cost" is just duplication with extra steps.

Can AI actually read your website?

We have moved into a different era of searching, and most websites have not caught up.

We still type the question into a browser, but increasingly we do not get a list of blue links to work through; instead, we get an answer, assembled from sites the model could read and decided to trust. Which means before anything can recommend you, it has to be able to read you at all.

Here are four things that decide whether it can:

Your site has to let AI in. Most do by default. Some block it, sometimes without the owner knowing, and a blocked site is not ranked lower - it is absent.

Your words have to be actual text. This one catches more people than you would think. Your services, your prices and your whole value proposition set beautifully inside a graphic are unreadable to AI.

Answer questions directly. Ask the question, then answer it plainly, in a sentence or two. Three paragraphs of setting the scene first is how you get skipped.

It has to be able to tell who you are. Your name, your credentials, your reviews, and your business details should say the same thing everywhere they appear. Models are trying to resolve you as one identifiable thing, and inconsistency reads as uncertainty.

There is a fifth one nobody enjoys hearing: you get cited more when other places already talk about you. Directories, profiles, other people's articles, forum threads are the place to be. Your own website is not enough on its own; you need backlinks to get found.

The unglamorous basics still count

Page titles, headings, meta descriptions, and a site that loads quickly are still doing the heavy lifting in the background. Especially the loading speed of the site, since nobody wants to wait five seconds to find out whether you are any good.

One correction, because this is the single most repeated piece of wrong advice in the category: your meta description is not a ranking factor. Google has said so plainly and repeatedly. What it does is persuade somebody to click once you are already showing up, and it could be the difference between getting the click or getting scrolled past.

However, your title tag is a different matter. That one still counts, and I mean alot!

Google is not the only search engine anymore

The last tip has nothing to do with your website. People search on YouTube, TikTok, Reddit and Pinterest now, and they often trust what they find there more than a page that is obviously trying to sell them something. A Reddit thread where somebody honestly answers a question carries a weight that a services page structurally cannot.

Which means a well-made Instagram post is SEO. So is answering a question properly in a forum where your clients actually are, and so is a Pinterest pin that is sending you traffic two years after you made it.

Pick one platform besides Google and genuinely show up there. One is enough, and one done properly beats five accounts posting the same thing into the void.

What I am deliberately skipping

You must have seen llms.txt recommended as the new essential file, possibly several times. It is a text file you add to your site to tell AI systems what matters on it, and the idea is genuinely appealing.

But here is the thing: no major AI service has said that it uses the file, and server logs show the crawlers do not request it. Even Google's John Mueller compared it to the old keywords meta tag, which is about as damning as that comparison gets. Any guide still recommending it in late 2026 is recycling advice from early 2025.

I mention it because the desire for a single file that fixes everything is exactly how the last decade of bad SEO advice spread. There is not one thing; it's many things done consistently that get results.

What happens next

I will report back. That is most of the point of doing this in public: in a few months I will have real numbers from my own site rather than borrowed ones, and I would rather show you what actually moved than what was supposed to.

Currently, advanced SEO is a service I offer to my web design clients. But it is something I am deep in, and it is very likely coming soon as a stand-alone service.

Frequently asked questions

Does SEO still matter if AI is answering people's questions?

More than before, not less. AI answers are assembled from sites that could be read and were trusted. The work that makes you findable in search is largely the same work that makes you quotable in an AI answer.

How many pages should a small business website have?

For most service businesses that is a home page, one page per service, an about page, a contact page and a blog. Pages you add without a reason become pages nobody visits.

Is it worth adding FAQs now that Google removed FAQ rich results?

Yes, but for a different reason than before. You no longer get the expandable dropdowns in search results. But you do get content in exactly the shape AI systems extract from, and questions your readers genuinely wanted answered.

Do meta descriptions affect Google rankings?

No. Google has confirmed the meta description is not a ranking factor. It affects whether somebody clicks your result once you are already ranking, which matters, but it will not move your position.

How long does SEO take to work?

Months, not weeks, and longer on a new site than an established one. Anybody promising results in thirty days is either redefining "results" or selling something.