Where to Actually Start With SEO on a New Site

Ask what the first step in SEO is and most answers start with keyword research. That's a reasonable second or third step. It's the wrong first one, because keyword research on a site that search engines can't crawl, index or tell apart is work you'll have to redo.

The order that actually matters

Before a page can rank for anything, four things must be true in sequence:

1. A search engine can reach it.
2. It's allowed to index it.
3. It can tell what the page is about.
4. It can tell it apart from your other pages.

Each depends on the one before. A brilliantly written page that returns a 404 to crawlers ranks for nothing. This is unglamorous work, mostly done once, and it's genuinely the first step.

Step one: can it be reached?

A page needs a path to it. Three things to verify:

• It's linked from somewhere. A page with no internal links pointing at it — an "orphan" — may never be discovered. Every page should be reachable by following links from the homepage.
• It returns HTTP 200. Sounds obvious; broken internal links and pages that error only for logged-out visitors are common. Check as an anonymous visitor.
• There's a sitemap. An XML sitemap listing your real URLs, submitted in Search Console. It doesn't guarantee indexing but removes the excuse of not knowing a page exists.

If your site depends on JavaScript to render content, this step needs more care — the crawler must execute that script to see anything. Server-rendered HTML avoids the question entirely.

Step two: is it allowed to be indexed?

The failure mode here is self-inflicted and surprisingly common: blocking your own site.

• robots.txt — controls crawling. A Disallow: / left over from staging blocks the entire site. This is the single most damaging one-line mistake in SEO.
• <meta name="robots" content="noindex"> — controls indexing. Correct on search results and thank-you pages; catastrophic if a template applies it site-wide.

Step three: can it be described?

Three elements do most of the work, and all three should describe the page honestly rather than being stuffed:

• Title tag — the strongest on-page signal, and the headline in results. Unique per page, front-load what the page is about, keep it around 60 characters before truncation.
• Meta description — not a ranking factor, but it's the sales copy under your title. Without one, a search engine excerpts something arbitrary. Around 155 characters.
• One H1 that matches the page. A single top-level heading stating the subject, then H2s for real sections in logical order — not chosen for visual size.

Structured data (schema.org markup) fits here too. It doesn't directly raise rankings, but it enables richer result formats — dates, breadcrumbs, FAQs — that affect whether anyone clicks. Add it once the basics are right, not before.

Step four: is it distinguishable?

The same content reachable at several URLs splits its own signals and forces a search engine to guess which is authoritative.

Two fixes, and both are worth doing:

• A canonical tag on every page declaring its one true URL.
• Consistency at the source — pick one form (with or without www, HTTPS only, trailing slash or not) and redirect the alternatives rather than serving both.

This site does exactly that: every content item has one public URL, and internal renderer routes redirect to it rather than serving duplicate copies. Canonicals are generated from the same source that generates the links, so they cannot disagree.

Step five: is it usable?

Two things here are genuinely part of the technical baseline:

• Mobile. Indexing is mobile-first — the mobile rendering is what gets assessed. If it's cramped or unreadable on a phone, that's the version being judged.
• Speed and stability. Core Web Vitals — LCP under 2.5s, INP under 200ms, CLS under 0.1. A modest ranking factor, but a substantial factor in whether people stay; see Why Visitors Leave.

HTTPS belongs in this list too, and is now table stakes rather than an advantage.

Only then, content and links

With the above done, the remaining work is the part everyone wanted to start with — and it now has a foundation to stand on: understanding what people are actually searching for (see Keywords in SEO), writing pages that genuinely answer those questions better than what currently ranks, and earning links by being worth linking to.

That last point deserves a caveat the old version of this article got wrong: link acquisition schemes — directory submissions, guest-post swaps, paid links — range from ineffective to penalised. Links that help are the ones someone chose to give you.

Most of the technical items above appear as checks in the Website Launch Checker, which is a faster way to confirm you haven't missed one.