SEO: A Complete Guide to How Search Engine Optimisation Actually Works

Search engine optimisation has an image problem. It is sold as a set of tricks, discussed as if it changed fundamentally every few months, and blamed whenever traffic falls. The reality is duller and more useful: search engines are trying to answer questions well, and SEO is the work of being the best available answer while remaining findable.

This guide covers the mechanism — what search engines do, what they reward, and what the work consists of. It is deliberately foundational; if you are looking for prioritisation and execution sequence, that is a separate matter covered elsewhere on this blog.

What a search engine actually does

Three distinct stages, and confusing them causes most diagnostic mistakes.

Crawling. A program follows links and requests pages. It has a finite budget per site. If your pages are unreachable, blocked, or reachable only through a form or a script, they are not crawled and nothing downstream can happen.

Indexing. The crawled page is understood and stored — or not. A page can be crawled and still not indexed, because it duplicates another page, because it has too little substance, or because the engine judged it not worth keeping. "It is in Google" and "Google has fetched it" are different claims.

Ranking. For a given query, indexed pages are ordered. This is the stage everyone talks about and the last one that matters — a page that is not indexed cannot rank no matter how good it is.

What ranking actually weighs

Nobody outside the engines knows the full recipe, and anyone claiming a precise list is guessing. But the broad categories are well established and stable.

Relevance. Does the page address the query? Not merely contain the words — address the need behind them. Someone typing "how much does X cost" wants a number, not a request for a quotation.

Quality and evidence. Is the page substantial, accurate, and produced by someone with a reason to be believed? This matters most where the stakes are high — health, money, legal exposure — and least for trivia.

Authority. Do other credible sites reference this one? Links remain a significant signal, but their value comes from being editorially given. Bought links carry risk without meaningful upside.

Experience of the page. Speed, mobile usability, stability, and whether the content is reachable without dismissing three overlays. These rarely make a bad page rank; they routinely stop a good one.

Context. Location, language and history alter results substantially. There is no single ranking — there is a ranking for a searcher in a place, in a language, on a device.

The three areas of work

Technical. Making the site crawlable, indexable and fast: one URL per page, correct canonicals, sitemaps listing only real pages, structured data that matches the content, and a mobile experience that works on a poor connection. Technical work rarely creates growth by itself; it removes the ceiling on everything else.

Content. Having a page for each thing people search for, written from the query rather than from internal vocabulary, specific enough that a competitor could not republish it with their name swapped in. This is where most durable gains come from, and where most effort is wasted on generic material.

Authority and reputation. Being referenced, reviewed and cited by sources with credibility. Slow, largely outside direct control, and the reason established sites outrank better pages from unknown ones.

Search intent, which decides everything else

The same words mean different things. "Running shoes" might be someone researching, comparing or buying. Search engines infer the dominant intent from behaviour and shape the results accordingly — which is why some queries return shops, some return guides, and some return a map.

What has genuinely changed, and what has not

Two real shifts are worth understanding.

Search engines have become far better at meaning than at matching. Writing the same phrase fifteen times stopped working long ago; covering a topic properly, including the questions that surround it, works well.

AI-generated answers increasingly sit above the results, which changes the shape of some traffic — particularly simple factual queries, where the answer is given without a click. What survives this is content offering something a summary cannot: judgement, specific experience, current local knowledge, evidence.

The mistakes that account for most failures

• Optimising for a keyword nobody uses. Check the words customers actually type before building a page around a phrase from your brochure.
• Multiple pages for one intent. They compete with each other and none of them wins.
• Confusing crawling with indexing. Leading to months spent on ranking work while the pages were never indexed.
• Judging by rankings alone. Position is a diagnostic; enquiries are the result.
• Redesigning without redirects. The single fastest way to discard years of accumulated visibility.

How to tell whether it is working