How I'd start a content site from scratch today (2026)

People ask me how they’d build a content site if they were starting now, with everything I’ve learned across 20 years and more sites than I’d admit to. So here’s the honest answer: the exact order I’d follow today, in 2026, knowing what actually matters and what’s a waste of the first year. It’s not clever. It’s a sequence — and doing it in order is most of the advantage.

1. Pick a niche I can genuinely own

I’d start narrow — narrow enough that one person can plausibly become the authority, not a faceless also-ran. The test: is there real search demand, a way to monetise it, and an angle I can cover better than a content farm because I’ve actually done the thing? Broad niches are where solos go to lose. A specific audience, region or use-case is winnable. (My bias is the EU-first, real-operator angle — the gap nobody fills well.)

2. Research before I write a word

I’d map the queries and their intent first — the actual things people search, and whether they want a guide, a comparison or a product. I’d target the winnable long-tail, not vanity head terms, and group queries into clusters. Writing without this is how you produce articles nobody searches for.

3. Build clean, fast and simple

I’d build on a fast, technically sound, low-cost stack — these days I can vibecode a clean site in a weekend. No bloated theme, no fighting a heavy CMS; just a quick, crawlable, mobile-fast site so the technical foundations never cap the content. The build is the easy part now — I’d spend the time on the data and the words, not the HTML.

4. Publish in clusters, not scattered posts

I wouldn’t publish random articles. I’d build topic clusters: a pillar page on a core topic, supported by spoke articles on its sub-topics, all interlinked. This is how a solo builds topical authority on a tight set of topics instead of being thin across many — and it’s exactly how this site is built.

5. Layer real experience through everything

The one thing a faceless AI content farm can’t fake is genuine first-hand experience — real numbers, real screenshots, real “here’s what happened when I did this.” I’d weave it through every piece. It’s the strongest, most defensible E-E-A-T lever a solo has, and increasingly the thing both Google and AI engines reward.

6. Write for AI search from day one

I wouldn’t treat GEO as an afterthought. I’d write answer-first, quotable, well-structured content so AI engines can cite it — because being the source an AI quotes is the freshest, least-crowded opportunity going, and it costs nothing extra to write that way from the start.

7. Wire monetisation in early — then be patient

I’d decide how it makes money before I need it — affiliate built into reviews, or my own product — wired in from the start, not bolted on later. And then the hardest part: I’d keep going through the long, quiet middle where nothing visibly happens, because content compounds and almost everyone quits right before it does.

What I would not do

• Chase a broad niche to compete with everyone — narrow wins for a solo.
• Publish scattered posts with no cluster or internal-link structure.
• Skip the foundations to rush content onto a slow, messy site.
• Wait to “feel ready” — start, publish, and let it compound; the asset is built in public.
• Quit in the quiet middle — the part where it looks like failure is the part where it’s working.

The takeaway

• A content site is a sequence: niche → research → build → clusters → experience → AI-ready → monetise → patience.
• Pick a niche you can genuinely own, with demand and a way to earn.
• Research intent first, build clean, and publish in clusters for topical authority.
• Real experience is the moat AI can’t fake — and AI-search readiness is free if you write for it from the start.
• Wire monetisation in early, then outlast the quiet middle — that patience is the whole game.

If I handed my younger self one page, it’d be this: stop looking for the clever shortcut. The boring, ordered version — done consistently while everyone else quits — is the shortcut.