SEO basics: what actually moves, and what is folklore

Search engine optimisation attracts folklore because the system it describes is opaque, changes, and produces correlations that look like causes. The discipline is to separate the parts you control from the parts you are guessing about.

The parts that are not in dispute

Across the published positions of the major engines, the same four things matter, and they are not technical.

1. Is the page about the thing searched for? The content has to answer the query, visibly and early. A page that answers a different question and hopes to rank for this one does not.
2. Can a machine read it? Text that renders, headings that describe structure, a title that says what the page is. This is the genuine technical part, and it is smaller than the industry around it suggests.
3. Does the site deserve to be trusted on this topic? A site with one good page on a subject and fifty unrelated ones is less credible on that subject than a site that is about it.
4. Do people who arrive stay and do something? Behaviour signals are contested in their weighting but not in their direction, and a page people immediately leave serves nobody, including you.

The folklore

• Keyword density. Repeating a phrase more often does not help and eventually makes the page bad. The phrase belongs where a reader would expect it: title, heading, first paragraph.
• A magic word count. There is no length that ranks. There is a length that answers the question, and it varies.
• Meta keywords. Not used by major engines, and have not been for well over a decade.
• Submitting your site everywhere. A sitemap and a little patience do the same job.
• Internal link counts. Linking more does not help. Linking to the right page with the right words does.

The technical checks that do matter

A short list, in order of how often they are actually broken:

• Every page has a title that is specific and not a template.
• Every page has one top-level heading that matches its title closely.
• The site can be crawled, with no blocked paths and no accidental noindex.
• Pages load without making the visitor wait for scripts that fail.
• Content is reachable at one address, not five variations of it.

Anything beyond this list is worth doing only when something is demonstrably broken. Site speed, structured data and clean canonical addresses all matter, and the common error is doing them before there is content worth being found.

Topical coverage, which is the part that compounds

A single page on a topic competes with sites that have twenty. Covering a topic properly — the main question, the adjacent ones, the confusions, the comparisons — is what makes a site the obvious answer for the whole area rather than for one phrase.

This is the same argument as content that compounds: the pieces that keep working are usually the ones that belong to a set.

What to do when the ranking drops

Find out what changed before changing anything. A drop can come from your site, from a competitor improving, from a change in what the engine decides the query means, or from seasonality. Changing several things at once in response to an unexplained drop guarantees you will not learn which of them mattered.

The honest measurement of whether any of this worked is in measuring marketing honestly.

What the page needs before anything technical

Four items, in this order: a title that says what the page is about, a first paragraph that answers the query, headings that describe the structure of the answer, and a conclusion that tells the reader what to do. A page with these four is competitive even if nothing else is optimised.

The commonest failure is a page that begins with context. Context is what the writer needed to write; the answer is what the reader came for, and it belongs first.

Links, both directions

Internal links are the only ranking factor entirely under your control, and they are usually neglected in favour of external ones. Linking from a page that is already found to a page you want found is free, immediate and reversible.

The words in the link matter, because they tell the engine what the destination is about. A link reading “click here” passes no information; a link reading “how to substantiate a claim” describes the target.

Patience, and what waiting actually looks like

Search performance moves over months rather than days, and the shape of the curve is usually flat then rising rather than the reverse. A page that gets no traffic in fortnight one is not necessarily failing; a page that has not moved in six months probably is, and the fix is usually that it answers a question nobody asks rather than that it needs technical work.