Search Engine Optimization (SEO): How Pages Get Found

Search engine optimization is the practice of making content discoverable, accessible, and interpretable to a search engine's automated systems, and relevant enough to the need behind a query, in order to earn unpaid search visibility.

Optimization

Search engine optimization, or SEO, is the work of making content that a search engine can find, fetch, read, store, and then judge relevant enough to show someone who is searching. That is the entire discipline. Most disappointment with SEO comes from treating it as one lever that produces rankings, when the system underneath is a sequence of stages, each able to drop a page before anyone sees it.

Nearly all evidence below is Google's own documentation. Where a claim is Google's, this entry says Google rather than search engines.

SEO, Defined by What the Search System Does

A durable definition describes the system, not the service. Google describes Search as a fully automated system that finds pages with crawlers, stores what it understands in an index, and returns results programmatically. Google also states that it does not accept payment to crawl a site more often or rank it higher.

Paid placement is bought through an advertising product and labeled as such. Organic placement is decided by software and cannot be purchased. SEO influences the inputs to that software, not the outcome.

The Gates Between Publishing and a Click

Google's guide to how Search works describes three stages: crawling, indexing, and serving. Google states that not all pages make it through every stage. Treating them as gates shows where a visibility problem actually sits.

Discovery and Crawling

Google finds URLs by extracting links from pages it knows and by reading submitted sitemaps. There is no central registry of web pages. Googlebot then decides algorithmically which sites to crawl and how often, and does not crawl everything it discovers. Blocked paths, login walls, server errors, and robots.txt rules are named reasons a discovered URL never gets fetched.

Rendering happens here, inside the crawl. Google runs the JavaScript it finds using a recent version of Chrome. Practitioner diagrams often show crawl, render, index, and rank as four boxes. Google's wording describes three stages, with rendering inside the crawl.

Indexing and Canonicalization

During indexing, Google analyzes page text along with key tags such as title elements and alt text. Indexing is not automatic. Google names low content quality, robots meta rules, and site designs that obstruct indexing as reasons a processed page is never stored.

Canonicalization happens here, and it is the gate most often missed. Google groups pages with similar content into a cluster and picks the one it considers most representative. That page becomes canonical. The others become alternates. A site can pass every technical check and still lose a page to another URL inside its own cluster.

Retrieval, Ranking, and Serving

When a query arrives, Google searches the index and orders results using what it describes as hundreds of factors, which can include the searcher's location, language, and device. There is no published list, so a numbered set of ranking factors presented as fact contradicts the primary source.

Serving is a gate again. Google states that a page reported as indexed can still fail to appear, naming three causes: content not relevant to users' queries, low content quality, or robots meta rules preventing serving. Indexing is not a promise of serving.

Requirements, Recommendations, and Claims That Are Neither

Most confusion in SEO comes from collapsing different statements into one word: factor. Google publishes exactly three technical requirements. Googlebot must not be blocked. The page must return a success status. The page must have indexable content. Google attaches its own caveat, noting that meeting all three still does not mean a page will be indexed. Everything beyond them is a recommendation, confirmed system information, an outside observation, or a hypothesis.

How to Classify Common SEO Claims

This table separates SEO claims into requirements, recommendations, confirmed system information, measured observations, practitioner hypotheses, and unsupported claims.

Claim	Type	What the evidence supports	What it does not prove
Googlebot access, success status, indexable content	Explicit requirement	Published as the minimum for eligibility in Google Search	That meeting them gets a page indexed or served
Internal links that make content findable	Official recommendation	Listed among the SEO fundamentals that stay worthwhile	That a number of internal links improves position
Core Web Vitals	Confirmed system information	Google states there is no single page experience signal, and that Core Web Vitals are used by its ranking systems	That good scores produce higher rankings, or that other aspects directly help rank
Structured data	Official recommendation	Markup can make a page eligible for specific rich results	That it raises ranking, or that AI features require it
External links	Confirmed system information	Described as a discovery path and a relevance signal	That link counts or vendor scores map to how Google ranks
Zero-click rate of 68.01%	Measured observation	One US panel measured this, January to April 2026	That the rate applies elsewhere, or that AI features caused it
Dwell time as a ranking metric	Practitioner hypothesis	Nothing. The term has no first-party definition	Anything about how Google measures time on a page
An ideal keyword density	Unsupported claim	Nothing. No published source states a target	That any density figure affects ranking

The Working Areas of SEO

Conventions of labor, not categories Google publishes, each maps to a gate above.

Content, On-Page Relevance, and Search Intent

This covers what a page says and how clearly it says it. Google's bar is framed around uniqueness rather than formatting mechanics, favoring first-hand experience over material that restates common knowledge. Relevance is a match to the need behind a query, not its wording. Google warns against producing separate content for every variation of how people might search, and says doing so to influence rankings violates its scaled content abuse policy.

Technical SEO

Technical work decides whether the pipeline can reach, fetch, render, and store a page. Server responses, robots rules, crawl access, JavaScript handling, and site moves sit here. The floor is genuinely binary, and above it returns fall away quickly.

Architecture, Internal Links, and External Links

Internal architecture is a discovery mechanism first, since Google finds new pages through links extracted from known pages. External links are treated as both a path to new content and a signal in judging relevance. Google publishes no model of how link value is calculated, so quantified link claims belong in the hypothesis column.

Structured Data, Images, and Video

Structured data is an eligibility mechanism. Correct markup can make a page eligible to appear as a specific result type, which differs from ranking higher. Google notes that a site already following its image and video documentation is prepared for generative AI search.

What SEO Does Not Guarantee

Five words get used as if they mean the same thing. Eligibility means a page is allowed to be considered. Visibility means it appeared. Ranking is where it appeared. Traffic means somebody clicked. A business result is what the click produced. Every step between them is a gate.

Google states in three separate current documents that meeting every requirement, best practice, and policy still does not mean it will crawl, index, or serve a page. On its guidance for hiring an SEO, Google says nobody can guarantee a number one ranking, and treats that promise as a warning sign. Timelines are not fixed either, since Google states recrawling can take several days to several months. Readers weighing outside help can check a provider's stated scope, such as Arfadia's search engine optimization service.

What AI Search Changes, and What It Does Not

How Google's AI Features Select Sources

Google documents that its generative AI features are rooted in its core Search ranking and quality systems. Retrieval-augmented generation, which Google calls grounding, relies on those ranking systems to pull relevant pages from the index, then reviews them to generate a response with clickable supporting links. Query fan-out issues several related searches across subtopics at once, drawing on a wider set of pages than one classic search surfaces.

Eligibility is the practical part. To appear as a supporting link in AI Overviews or AI Mode, a page must be indexed, eligible to be shown with a snippet, and meeting the Search technical requirements. Google's AI features documentation states there are no additional technical requirements. Its newer guide adds a condition that is easy to miss: the site must be included in Search generative AI features in Search Console, which is the default state. Google notes AI Overviews appear only when its systems judge them additive, and often do not trigger.

What You Do Not Need to Do

Google's guide to optimizing for generative AI features lists things site owners can ignore for Google Search. Google Search does not use llms.txt or similar files, and keeping one will neither help nor harm visibility there. There is no requirement to chunk content, and Google states there is no ideal page length. Rewriting pages specifically for AI systems is unnecessary. Seeking inauthentic mentions is listed as unhelpful. Structured data is not required, and no special schema markup exists for it.

Where GEO and AEO Sit

Google addresses both labels by name. Its position is that from Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO. Two limits matter. That describes Google's view of its own surfaces, and it says nothing about optimizing for AI assistants that are not Google Search. Arfadia sets out its own scope on its generative engine optimization service page.

Measuring SEO When Visibility and Clicks Separate

Measurement used to end at rankings and sessions. It cannot now, because appearing and being clicked have come apart. Sites appearing in AI features are counted in overall Search Console traffic within the Web search type. Google also launched a dedicated generative AI performance report, which reached all websites worldwide on 31 August 2026 and groups data by pages, countries, dates, and devices.

The limit is specific. That report shows impressions only. There are no clicks, no click-through rate, no average position, and no queries in it. A site owner can now answer whether they are visible inside AI features. They cannot answer, from first-party data, what that visibility earned. Arfadia tracks that second question separately in its AI citation rate report.

The click evidence has to be read by metric rather than by headline. Pew Research Center tracked 900 US adults during March 2025 and found users clicked a traditional result on 8% of visits where an AI summary appeared, against 15% where none did. SparkToro, using Similarweb clickstream data for US searches from January to April 2026, reported 68.01% of Google searches ended with no click of any kind, with mobile in-app searches excluded. Google has stated that total organic click volume stayed broadly stable year over year, without publishing figures alongside the claim.

They are compatible, for a definitional reason. Rate and volume are different things. A click rate per search can fall while absolute clicks hold steady, if total search volume grows. None of this establishes that AI features caused a traffic change for any site, because all of it is associative. The figures are also US figures, from different instruments and periods, so they cannot be averaged or carried to another country.

A Worked Example From Indonesia

Google published a case study on Vidio, an Indonesian over-the-top video platform, which implemented VideoObject markup across its existing catalog and gave each video file a stable URL so Googlebot could reach it. Google reports roughly three times the video impressions and close to double the video clicks within a year, comparing the first quarter of 2022 with the same quarter in 2023.

Google's own write-up also states the number of videos Vidio published grew by around 30% across that window. Part of the impression growth is inventory growth, and the study does not separate the two. It illustrates eligibility and discoverability. It is not evidence of a ranking boost.

SEO in the Indonesian Market

Statcounter put Google at roughly 93% of tracked search referrals in Indonesia in August 2026, with other engines in low single digits. That figure comes from page views on sites carrying Statcounter's code, so the direction is reliable while the decimals are not. DataReportal counted about 230 million internet users in October 2025, around 80.5% of the population.

What does not exist is worth saying too. No usable Indonesia-specific measurement of zero-click behavior or AI Overview prevalence surfaced during research for this entry. The absence is itself a finding. Applying US panel results to Indonesian searchers would produce a number that looks authoritative and means nothing. For sourced Indonesian market context, Arfadia publishes a state of SEO Indonesia report and a digital marketing benchmark for Indonesia.

Frequently Asked Questions

What is SEO in simple terms?

SEO is the practice of making content a search engine can discover, access, interpret, and judge relevant to a search, so it can appear in unpaid results.

How does SEO actually work?

It improves conditions at each stage of a pipeline the site owner does not control. Google describes crawling, indexing, and serving, with rendering during the crawl and canonicalization during indexing. A page must pass every stage before anyone can click it, and an early-stage problem cannot be fixed by later effort.

How is SEO different from paid search?

The difference is mechanism, not screen position. Paid placement is purchased through an advertising product and labeled as sponsored. Organic placement is determined programmatically, and Google states it does not accept payment to rank higher.

Can SEO guarantee a number one ranking?

No, and Google says so directly. Its guidance on hiring an SEO states nobody can guarantee a number one ranking, and treats that promise as a warning sign. Google also states crawling, indexing, and serving are not guaranteed.

Do backlinks still matter?

Links matter qualitatively. Google describes them as a way new pages get discovered and as a signal in judging relevance. Google publishes no model of how link value is calculated, so claims tying a link count or vendor score to a ranking are unsupported.

Does AI Search change what SEO means?

It changes the presentation layer and the measurement picture more than the work. Eligibility for AI Overviews and AI Mode is normal indexing plus snippet eligibility plus the technical requirements, with site inclusion set in Search Console. No special files, markup, or chunking are needed.

How should SEO performance be measured?

Start from the objective, then work down: impressions, position indicators, organic clicks, on-site engagement, and the business outcome. Search Console is the first-party surface, including the generative AI performance report, which shows impressions rather than clicks. Arfadia also publishes an SEO ROI calculator.

Related Terms

Generative Engine Optimization (GEO), Technical SEO, Structured Data, Search Intent, Search Console.