SEO (search engine optimization)
The practice of making a page understandable and retrievable by search systems — defined by Google as helping engines understand your content and helping users decide whether to visit.
Karl-Gustav Kallasmaa, Founder & CEOLast updated Search engine optimization is the practice of making a page understandable and retrievable by a search system. Google's own SEO Starter Guide defines it as "helping search engines understand your content, and helping users find your site and make a decision about whether they should visit your site through a search engine" — a definition about comprehension and decision support, containing no promise about position.
The three layers, and what each actually controls
Google Search Essentials splits the requirements into three categories, and the distinction between them is the most useful thing in the discipline:
- Technical requirements — what Google needs from a page to show it in results at all. Reachable by a crawler, not blocked, returns a successful status, contains indexable content. This is binary. Fail it and nothing else on the page matters.
- Spam policies — behaviours that lead to lower ranking or complete omission. Also close to binary, in the wrong direction.
- Key best practices — the things that help how a site appears. This is advisory, probabilistic, and the layer that most SEO advice is about.
Most work sold as SEO operates in layer three while the actual failure sits in layer one. A page that renders its content client-side and times out, or that is behind a login, or that returns a soft 404, has a technical fault no amount of heading structure repairs.
The mechanism
A search system does three separable things: it discovers a URL, it fetches and parses the document, and it selects among what it has. Crawling and indexing is the first two; ranking is the third. Optimization work maps onto them cleanly:
- Discovery — internal links, XML sitemaps, and not blocking the fetch. A URL nothing links to and no sitemap declares may simply never be seen.
- Parsing — server-rendered HTML, one canonical URL per piece of content (canonical URLs), and explicit machine-readable claims via schema markup.
- Selection — relevance to the query, plus the page-experience signals Google publishes. The Core Web Vitals thresholds are documented: LCP within 2.5 seconds, INP of 200 milliseconds or less, CLS of 0.1 or less, each assessed at the 75th percentile of page loads segmented across mobile and desktop.
Google is explicit that feedback is slow: some changes take effect in a few hours, others several months, and the guide advises waiting a few weeks before judging. Any workflow that treats SEO as a same-week control loop is misreading the system it is operating.
What transfers to AI answers, and what does not
The technical layer transfers completely. A retrieval step inside an AI assistant is an HTTP fetch by a named agent, and it has the same preconditions as a search crawler: reachable, fast, unauthenticated, parseable. OpenAI documents distinct agents for distinct jobs — OAI-SearchBot "used to surface websites in search results in ChatGPT's search features", ChatGPT-User for page visits triggered by a user's question, and GPTBot for model training. A robots.txt rule written years ago against generic crawlers can silently exclude a site from the first two while the operator believes they have only opted out of training.
The ranking layer does not transfer. There is no position to hold in a generated paragraph. What is selected is a passage, and the qualities that make a passage selectable are different from the ones that make a page rank: one claim per section, the qualifier in the same sentence as the claim, identifiers written out literally so a lexical matcher can find them, and headings phrased as the question being answered.
Failure modes
- Optimising layer three over a broken layer one. Heading structure on a page a crawler cannot render is decoration.
- Treating a definition as a promise. Google's definition covers understanding and decision support. Vendors who describe SEO as a lever on position are describing an outcome, not a mechanism.
- Short measurement windows. Judging a change inside a week, against a system whose own documentation says the effect may take months, produces conclusions that are noise.
- Blocking AI agents by accident. Broad
Disallowrules or bot-management defaults exclude retrieval agents as readily as scrapers. See robots.txt. - Keyword stuffing. It degrades both classical ranking and passage extraction, because a passage packed with repeated terms states no single extractable claim.
Frequently asked questions
What does Google say SEO is?
Helping search engines understand your content, and helping users decide whether to visit — not a mechanism for position.
How fast do changes take effect?
Hours to months, per Google's own guide; assess after a few weeks, not a few days.
Do the technical requirements apply to AI assistants?
Yes. Retrieval is an HTTP fetch, with the same preconditions.
Does ranking well mean being cited?
No. Ranking places a document; citation selects a passage.
Terms related to SEO (search engine optimization)
The two separate stages that decide whether a page can be retrieved at all, and the reason a serving rule on a blocked page is never read.
The Schema.org vocabulary and the JSON-LD you embed with it, what it is genuinely good for, and the limits of what it can make happen.
Google's three published field metrics for loading, responsiveness and visual stability, with the exact thresholds and the percentile they are judged at.
The one URL you nominate as the indexable version of a page, how the rel=canonical hint is actually weighted, and what a split canonical costs you when an agent needs a link to cite.