What is Search Engine Optimization (SEO)?

At a glance

• SEO covers three areas: technical access, content that matches intent, and trust signals from elsewhere on the web.
• The output is a position on a results page, measured by rank, clicks and impressions.
• AI search has deprecated none of it. The foundation is shared and the scoreboard has doubled.
• The parts of SEO that age worst aimed at gaming a ranking rather than answering a question.

Crawl, index, rank: the three steps behind a listing

A search engine has to find your page, understand it, and place it against everything else on the same subject.

Crawling is the fetch. A bot requests your URL and reads what the server sends back. If the content appears only after JavaScript runs, or robots.txt blocks the path, this is where you fail.

Indexing is the storage. The engine parses the page, works out its subject, and files it. Duplicate pages, thin pages and pages with no clear subject get filed badly or not at all.

Ranking is the comparison. For a query, the engine orders the indexed pages it thinks answer it, weighing relevance, quality and trust. Google’s own starter guide sets out the basics honestly, and most of what an agency sells sits inside it.

Everything else in SEO is detail hung off those three steps.

SEO did not stop working, it stopped being the whole job

Search still sends more traffic than assistants do in almost every category. Anyone telling a marketing team to stop doing SEO is selling something.

What changed is that a good ranking no longer guarantees a visit. An AI Overview at the top of the page answers plenty of questions outright, so the click you used to get from position three sometimes never happens.

The second change is the second scoreboard. Your page can rank well and never get cited in an assistant’s answer, because AI search ranking picks passages on clarity and directness rather than domain strength.

So the honest position: SEO is necessary and no longer sufficient. The technical work it produces is what an AI crawler needs too, and its content standards are now enforced harder by a system that skips any page it cannot read cleanly.

The SEO work that still pays in an AI answer

Prioritize the parts of the discipline that make a page genuinely better, and both scoreboards move together.

1. Fix access before anything else. Server-rendered HTML, a correct sitemap, nothing important behind JavaScript, and a robots.txt that does not quietly block AI crawlers alongside search bots.
2. Write one page per real question, not per keyword variant. Consolidate the near-duplicates you built for phrase coverage, because both systems handle synonyms now.
3. Put the answer first. A direct sentence at the top serves the snippet, the reader and the assistant looking for something quotable.
4. Name your authors and show the evidence. E-E-A-T signals matter to ranking and to whether a model treats you as a source worth crediting.
5. Keep pages current and date them. Freshness affects both systems, and an undated page loses a tiebreak it never sees.