What is SEO, and what still actually works

What SEO actually is

SEO is the work of making a search engine likely to show your page to somebody who is looking for what you offer. That is the whole definition. Everything else is detail about how.

It is not a trick and it is not a purchase. You cannot pay Google for a higher position in the ordinary results, and anybody offering to guarantee one is either buying ads or lying.

The confusing part is that the work splits into three completely different kinds of job, done by different people, on different timescales. Most of the frustration people have with this subject comes from being sold one of the three as though it were all of it.

How a search engine finds you, in three steps

A search engine does three things in order: it visits your pages, it keeps a copy, and it decides which copy to show. If any step fails, nothing after it happens.

It visits. An automated program called a crawler arrives at your site and reads pages. Nobody is at a keyboard. It follows links, and it looks at your sitemap if you have one.
It keeps a copy. What it read goes into the index, which is the searchable store of the web. A page that is not in the index cannot appear in results, no matter how good it is.
It ranks. When somebody searches, the engine picks from the index and puts the results in an order on a page called the SERP.

Almost every serious problem in this field is a step-one or step-two problem being treated as a step-three problem. Somebody spends a year on writing when their pages were never being read.

The three jobs, in the order that matters

SEO is technical work, then content work, then reputation work, and the order is not a preference. Each one is wasted while the one above it is broken.

Technical. Can machines reach, read and store your pages. Days of work, usually done once, and it either passes or it does not.
Content. Do your pages answer what people are actually asking. Ongoing, and the part with the most room for judgement.
Reputation. Do other sites refer to you. Slowest, least controllable, most expensive, and the hardest to fake convincingly.

If you only ever do the first two properly, you will be ahead of most of your competitors, because the third is where agencies prefer to spend your money and the first is where the actual problems live.

Technical: can a machine read you at all

The technical job is a small set of yes-or-no questions, and you can check most of them yourself in an afternoon.

Are you turning machines away? Add /robots.txt to your domain and read it. That file lists which automated visitors are allowed. Blocks placed there years ago outlive whoever placed them.
Is a page telling itself not to exist? A single line called noindex keeps a page out of results entirely. It is the most common cause of a page that vanished for no visible reason, usually left over from a site launch.
Does your text arrive with the page? View the page source and search it for a sentence you know is there. If the sentence is missing, your site is built to assemble its text in the browser, and machines that do not wait see very little. The fix is called server-side rendering and a developer will understand it immediately.
Do you have several addresses for the same page? Shops generate these constantly through filters and sorting. A canonical tag names the real one so your standing is not split across copies of itself.
Is it painful to use? Core Web Vitals measure how fast and stable a page feels. They are a small ranking factor and a large reason people leave.

Content: answer the question somebody actually asked

Good content for search is not clever writing. It is a page that answers one real question completely, in the words the person asking would use.

The most common mistake is writing in the language of your own industry. Your customers do not know your internal vocabulary. A page titled around what a buyer types will beat a better-written page titled around what your team calls it.

The second most common mistake is burying the answer. Put it in the first paragraph, then explain. This has always been good writing, and it is now mechanically important as well: the short answer at the top is what gets pulled into a featured snippet or an AI answer.

A useful habit is to write one page per question rather than one page per topic. Ten pages that each answer one thing outperform one page that mentions ten things, because a machine matching a question to a page can only match what a page is clearly about. Doing that consistently across a subject is what people mean by topical authority.

Reputation: being mentioned by other people

A backlink is a link to you from somebody else’s website, and it remains the clearest signal that other people think you are worth pointing at.

The words used in the link, called the anchor text, tell a machine what the destination is about. A link that says our pricing carries more meaning than one that says click here.

Buying links is against Google’s rules and has been for fifteen years. It also still happens constantly, which is why you will be told both that it is fatal and that everyone does it. What is reliably true is that bought links from sites with no relationship to your subject are the easiest kind to detect and the least useful kind to have.

The one lever here you fully control is the internal link: pointing your own pages at each other. It is free, it is immediate, and it tells machines which of your pages you consider most important.

How to tell whether it is working

Use Search Console. It is free, it is Google’s own, and it is the only place you can see the actual queries your pages appeared for.

Watch two numbers together and the story usually explains itself. Impressions count how often you were shown. Click-through rate is the share of those who clicked.

Impressions rising, clicks rising: it is working.
Both falling: a ranking or indexing problem. Start at the technical list.
Impressions steady, clicks falling: people are being satisfied without you, which is the AI answer story below.

Positions on their own are a poor measure and always were. Being third for something nobody searches is worth less than being tenth for something they do.

What AI answers changed, and what they did not

The technical work did not change at all. A site a search engine cannot read is a site AI search cannot read, for identical reasons, on the same day.

What changed is the top of the results page. AI Overviews answer many questions before anybody scrolls, which produces zero-click searches: the person got what they needed and never visited a site. Informational pages feel this hardest. Pages people use rather than read barely feel it.

This is where the field genuinely disagrees, so both readings: one camp says search traffic is structurally declining and the sensible response is to plan for less of it. The other says the traffic that left was low-intent anyway and buyers still click when they intend to buy. Both have evidence, and which is true for you depends entirely on whether your pages are read or used.

What you can safely ignore

Keyword density. Repeating a phrase a set number of times has not worked for over a decade and now actively reads as padding.
Third-party authority scores. Numbers like domain authority are invented by tool vendors, not by Google. They are useful for rough comparison and are not a thing anybody is ranking you on.
Meta keywords. A tag no search engine has used since the 2000s. It still appears in checklists.
Anybody guaranteeing position one. Nobody controls that. The guarantee is the tell.

Glossary

Every term used above, each defined without needing another definition to make sense of it.

SEO: SEO, or search engine optimization, is the work of getting a page of your website to appear high in a list of search results.
Crawler: A crawler is a program that visits web pages on its own and copies what it finds, so a search engine or AI search can use the contents later.
Index: The index is the searchable copy of the web a search engine keeps. If a page is not in it, that page cannot appear in results.
SERP: A SERP is a search engine results page: everything Google shows you after you type a query.
Sitemap: A sitemap is a file listing the pages on your site that you want machines to know about.
robots.txt: robots.txt is a plain text file at the top level of a website that lists which automated visitors are allowed in and which are not.
noindex: noindex is an instruction in a page's code telling search engines not to include it in results.
Canonical: A canonical tag tells a search engine which version of a page is the real one when several addresses show the same content.
Server-side rendering: Server-side rendering means the page arrives from the server with its text already in it.
Client-side rendering: Client-side rendering means the page arrives nearly empty and its content is built by code running in the visitor's browser.
Backlink: A backlink is a link to your site from somebody else's.
Anchor text: Anchor text is the visible, clickable words of a link.
Internal link: An internal link is a link from one page of your site to another page of your site.
Topical authority: Topical authority is the degree to which a site is seen as covering a subject completely rather than mentioning it once.
E-E-A-T: E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness, the qualities Google says its raters look for.
Featured snippet: A featured snippet is the short answer Google lifts from a page and shows in a box at the top of the results.
Impressions: An impression is one instance of your link being shown to somebody, whether or not they clicked it.
Click-through rate: Click-through rate is the share of the people shown your link who actually clicked it.
Search Console: Search Console is Google's free tool showing how your site performs in its search results, including which queries showed your pages.
Core Web Vitals: Core Web Vitals are Google's three measurements of how fast and stable a page feels to a real person.
AI Overviews: AI Overviews are the AI-written answers Google places above its normal list of links.
Zero-click search: A zero-click search is one where the person gets what they needed without visiting any website.

Next: what GEO is and how to do it, which covers the same buyer arriving through AI search instead.

See where your own site stands

Everything above is measurable. The scan reads your site the way AI search does and tells you which of these it already passes.