Being Found: SEO and AI Search
Search behaviour reflects intent. When people search, they are already motivated. The marketer's job is not to interrupt that intent but to meet it, with content useful enough to be surfaced. As search shifts from ten blue links to a single synthesised answer, being found stops meaning ranking first and starts meaning being trusted enough to be referenced.
The language you will use this week
Six terms. Read the term, predict the definition, then flip. The vocabulary of search is changing under our feet: the words that described a list of links do not describe an answer engine, and the difference is strategic, not technical.
From ten blue links to one answer. Classic search rewarded position: rank first and win the click. Answer engines synthesise a single response and cite a small handful of sources, so visibility shifts from ranking to being trusted enough to be referenced. The fundamentals, clear structure, relevance, and authority, carry across both worlds.
Original diagram, after the AI-search discussion in Clement (2026), Chapter 5, and Moz, Beginner's Guide to SEO.By the end of this chapter you will be able to
- Explain why search behaviour reflects active intent and why that makes it a high-value discovery surface.
- Describe the signals search engines use to decide what to surface: relevance, authority, structure, and speed.
- Recognise that platform-native search, such as Naver in Korea, follows local logic that Google-shaped strategies miss.
- Explain how AI answer engines change visibility from ranking high to being trusted enough to be referenced.
- Distinguish content written to win the click from content written to be cited, and say why the latter is more durable.
- Apply the RANKS rubric to audit a page's readiness to be found and referenced.
The Search Intent Quadrant
Informational, commercial, navigational, transactional — the same person searches differently at each stage. Match a keyword to its intent across five products and services before you decide what page to build.
Open the interactive →6.1 Search as intent
Search engine optimisation matters because search behaviour reflects intent. When users search, they are already motivated. They are not being interrupted by a message; they have raised their hand and asked a question. That makes search fundamentally different from broadcast advertising. Good SEO aligns content with the real question a person is asking, rather than forcing a message onto an uninterested audience.
The launch of Google in 1998 changed discovery by shifting marketing from broadcasting messages to appearing at moments of intent. That principle has not weakened; if anything, intent has become the most valuable thing a brand can meet. What has changed is the interface through which intent arrives, and the rules for being present when it does.
A query is a declaration of active demand
The person searching has already decided they want to make progress. The marketer's task is not to create demand but to meet it, with content useful enough that the engine, or the answer model, chooses to surface it. Visibility now depends on usefulness, not just optimisation.
Intent is not one thing. The same person searches differently depending on how close they are to acting, and each kind of query wants a different page. The map below sorts the four classic intent types and what meets each. Build your own row for a query your audience actually types.
| Query type | Example query | Intent | What to publish to meet it |
|---|---|---|---|
| Informational | "how to brew Ethiopian coffee" | Learn, not yet buying | A clear how-to guide, structured so a model can lift it |
| Navigational | "[brand] single origin" | Reach a known brand | A fast, clean brand and product page |
| Commercial | "best single-origin beans 2026" | Compare before buying | An honest comparison with concrete specifics |
| Transactional | "buy Ethiopian beans Tokyo" | Ready to act | A frictionless product page with local availability |
6.2 How engines decide what to surface
For most of search history, visibility depended primarily on ranking, and ranking depended on a handful of durable signals. They have not disappeared. Clear structure, relevance, and authority still matter. What has changed is that they are now table stakes rather than a winning hand.
| Signal | The question it answers | What weakens it |
|---|---|---|
| Relevance | Does this content actually answer the query? | Keyword matching without substance |
| Authority | Can this source be trusted on the topic? | No track record, no credible references |
| Structure | Can the content be read and parsed cleanly? | Walls of text, no headings, no clear answer |
| Speed | Does the page load fast enough to keep the user? | Heavy pages that lose the visitor before they read |
Still the default for most of the world
For a quarter of a century, Google has been the place intent arrives for most of the planet. Its long evolution, from counting links, to understanding meaning, to surfacing AI overviews above the results, is a single story: each step rewarded usefulness more and raw optimisation less. The brands that survived every algorithm update were rarely the ones who gamed it. They were the ones whose content genuinely answered the question better than the alternatives.
For marketers, Google remains the discipline that teaches the fundamentals. Relevance, authority, structure, and speed are not Google-specific tricks; they are what any system, human or model, uses to decide whether a source is worth surfacing. Master them on Google and they transfer everywhere else.
Moz. (2025). The beginner's guide to SEO. moz.com. Brand mark used for editorial reference under educational fair use.Clicks collapse below the first result. The top organic position takes roughly a third of clicks; by the fifth, almost none. Two forces now compress this further: answer engines that resolve the query with no click, and AI overviews that push the ten links down the page. Ranking first still matters, but it is no longer the whole game.
Illustrative of well-documented click-through patterns, not a single source. See Dataset 6.A below.| Position | Share of clicks | What it means |
|---|---|---|
| Position 1 | ~32% | Takes the lion's share of attention |
| Position 2 | ~17% | Already less than half of position one |
| Position 3 | ~11% | The long fall begins |
| Position 4 | ~7% | Below the fold for many users |
| Position 5 | ~5% | Rarely the first choice |
| Positions 6 to 10 | ~8% combined | Seen by few, clicked by fewer |
| No click | ~20% | Answered on the page or by an AI overview |
Reading a ranking signal
A page repeats its target keyword forty times but never actually answers the question a searcher asked. It ranks briefly, then drops. Which response best reflects this chapter?
C is correct. Optimisation without substance is fragile. Engines reward content that resolves the searcher's intent, not content that merely contains the keyword. Section 6.2.
6.3 Platform-native search
Search does not behave the same way everywhere. Channels are cultural artefacts, not universal tools. In Korea, platforms like Naver and Kakao dominate discovery, search behaves differently, and ecosystems are closed rather than link-based. A Western SEO strategy copied directly into this environment usually fails, because it optimises for a link graph that is not the one users actually move through.
Korean SEO is platform-native, not Google-shaped
A Naver results page does not look like a Google results page. It blends Naver Blog posts, Knowledge-iN questions and answers, Naver Shopping price tiles, Naver Place map results, and video into one layered surface. Trust signals are platform-internal: a credible Naver Blog review can carry more weight than a brand's own website, and Knowledge-iN answers shape decisions long before a user reaches a .com. Being found on Naver means living inside Naver, not pointing at it from outside.
For a brand entering Korea, the practical implication is concrete. Maintain a Naver Blog presence, seed honest Knowledge-iN answers, register with Naver Place, and treat Naver Shopping tiles as a primary surface. Global platforms do not erase local behaviour; they sit on top of it. The marketer who adapts to local platform logic beats the one who imports global assumptions.
Naver. (2025). Advertising and business resources. business.naver.com. Statista. (2025). Internet usage in South Korea. Brand mark used for editorial reference under educational fair use.The same query, two shapes of a results page. Google's page is a ranked list pointing outward to the open web. Naver's page is a blended surface, blog, shopping, Knowledge-iN, place, and video, that keeps the user inside Naver. A brand optimising only for the link-graph shape is invisible on the blended one.
Original diagram, after the platform-native search framing in Section 6.3 and Naver's public description of its search results layout.6.4 The decline of old SEO assumptions
Traditional SEO assumed that people searched with keywords and that visibility depended primarily on ranking. That logic is weakening. AI-mediated search systems increasingly interpret questions, summarise answers, and recommend sources directly. Users receive synthesised responses rather than lists of links. This changes what visibility means.
Being found is no longer only about ranking high. It is about being trusted enough to be referenced. Chapter 6 · 6.4
The citation-first answer engine
Perplexity treats the cited source as a first-class object. Every answer shows the small set of pages it leaned on, numbered inline, so the reader can see who the model trusted. The strategic consequence is that the payoff has moved from clicks to citations. A page that is the second or third citation in a comparison answer has done its job, even if the user never clicks through, because it has shaped the answer the user acts on.
This rewards a different kind of content. Citation-friendly pages are specific, structured, and willing to state facts plainly and compare honestly. Marketing copy without facts gets ignored, because there is nothing in it for the model to lift and attribute. Optimisation without substance is fragile in exactly the place it used to be safe.
Perplexity. (2025). How Perplexity sources answers. perplexity.ai. Brand mark used for editorial reference under educational fair use.None of this eliminates SEO fundamentals. Clear structure, relevance, and authority still decide whether a model considers a source at all. But content that genuinely explains, clarifies, or guides a decision is more likely to be surfaced than content designed only to attract clicks. The implication is strategic, not tactical: visibility now depends on usefulness.
The six tests of being found
- The page answers a real question someone is actively searching, not a message we want to push.
- It carries credible authority signals: experience, evidence, and references a model can trust.
- It is structured cleanly, with headings and a clear answer near the top, so it can be parsed and lifted.
- It loads fast enough that the visitor stays long enough to read it.
- In our primary market, we are present inside the platform-native search surfaces our audience actually uses.
- At least one page states concrete, structured facts designed to be cited by an answer engine, not just clicked.
6.5 Conversational search
The newest shift is that search is becoming a conversation. Instead of refining a query string, the user refines a dialogue: ask, read, follow up, narrow. The discoverable unit is no longer a page; it is a fact stated clearly enough that the model picks it up and repeats it correctly.
Conversation as the new query
Web-connected ChatGPT moved a meaningful slice of how-to and compare-X-versus-Y questions out of the results page and into multi-turn chat. The user does not scan ten links; they ask, read a synthesised answer, then ask again. For a brand, this means the smallest findable thing is a clear, correct, well-attributed fact. If the model can lift it cleanly and trust the source, the brand becomes part of the answer. If the fact is buried in marketing language, the model skips it.
The defensive lesson is uncomfortable for teams whose content was written to win a click. Conversational search rewards clarity and correctness over persuasion. The brand that explains best, not the brand that shouts loudest, gets repeated.
OpenAI. (2024). Introducing ChatGPT Search. openai.com. Brand mark used for editorial reference under educational fair use.Mozlow’s hierarchy of SEO needs. Crawl accessibility sits at the base and rankings sit at the top. You cannot earn the upper tiers until the lower ones hold.
Original diagram, after Moz’s hierarchy of SEO needs, in Clement (2026), Chapter 6.How content earns a citation. An answer engine narrows candidates through filters that echo classic SEO, relevance and authority, then adds two that punish marketing copy, structure and substance. Content written only to win a click tends to fail at the authority filter, where there is nothing for the model to trust and attribute.
Original diagram, after the AI-search framing in Clement (2026), Chapter 5.Writing for citation
A coffee brand wants its sourcing page cited by answer engines. Which change most increases the chance of being referenced?
C is correct. Answer engines favour content that is specific, structured, and directly addressable. The discoverable unit is a clear fact the model can lift and attribute. Section 6.5.
6.6 Where AI answers come from
If being found now means being cited, the blunt practical question is this: where do answer engines actually get their information, and which sources do they weight most? Two pipelines feed every AI answer. Training data is the frozen snapshot a model learned from. Live retrieval, often called RAG, is the small handful of pages the engine fetches at the moment of the question and quotes back. Shaping content so those systems pick it up has a name, Generative Engine Optimisation (GEO), and the striking finding is how few sources dominate both pipelines.
The concentration is stark. In one analysis of about 150,000 AI citations, Reddit was cited in 40.1% of answers and Wikipedia in 26.3%, with community and reference sites far ahead of any individual brand's own website. A separate 2026 index estimated that just 15 domains account for roughly 68% of all consolidated AI citations, and Ahrefs found only about 12% of AI citations overlap with Google's own top ten results. Being found by a model is therefore less about your page ranking first and more about being present, and well-regarded, on the platforms the models already trust.
The weights move under your feet. After a September 2025 change in how ChatGPT retrieved results, its citations of Reddit and Wikipedia, its two most-used sources, fell sharply within weeks. The lesson is not to chase one platform but to build durable presence across several, so a single shift cannot erase your visibility.
Semrush, 13-week study of 230,000+ prompts (2025). Figures approximate, read from the reported trend.Different engines trust different corners of the web, so "being found" is really several jobs at once. The table below sorts where the major engines lean and what each preference asks a brand to actually do.
| Engine | Leans toward | What that asks of a brand |
|---|---|---|
| ChatGPT Search | Wikipedia, Reddit, high-authority editorial such as Forbes | Encyclopedic, well-attributed facts and an earned Wikipedia presence |
| Perplexity | Reddit, LinkedIn, and primary sources such as NIH | Cite primary data and keep pages fresh; recency is rewarded |
| Google AI Overviews & AI Mode | Google's own properties (YouTube) and social or review sites (Facebook, Yelp) | Own a real YouTube presence; keep third-party reviews strong |
| Gemini | YouTube transcripts and Google-indexed content | Publish video whose spoken transcript states the facts plainly |
The pattern points to where a brand can honestly act. Notice that none of these are the brand's own website first: the highest-weight signals are earned on platforms the models already trust.
| Surface | Why models weight it | The ethical move |
|---|---|---|
| Reads as authentic, community-validated experience | Answer real questions in your category; add genuine value, never plant promotional posts | |
| YouTube | Transcripts are indexed and heavily reused, especially by Google engines | Earn mentions in honest reviews and tutorials; say the facts out loud on camera |
| Wikipedia | A foundational reference layer for factual answers | Earn the secondary coverage that qualifies you; never pay for edits, they get reverted |
| Rising fast for professional and B2B queries | Publish substantive founder and employee posts, not reposted marketing | |
| Review aggregators (G2, Trustpilot, Yelp, Tripadvisor) | Decisive in category-specific answers | Earn real reviews and keep listings accurate; do not buy or fake ratings |
Earn the citation, do not manufacture it
Because AI weights community and review platforms so heavily, the temptation is to game them: astroturf Reddit, buy reviews, pay for a Wikipedia entry, or seed manipulative text to steer a model. These backfire. Moderators and Wikipedia's editors reliably remove conflict-of-interest content, and researchers have shown that user-generated content can be poisoned to mislead AI, which is exactly why platforms and models are learning to distrust inauthentic signals. The durable strategy is E-E-A-T: real experience, expertise, authority, and trust, expressed as genuine participation and verifiable facts. What is easy to fake is easy for a model to learn to discount.
Two research-backed tactics survive that ethical bar because they make content genuinely more useful. Content that states concrete statistics is roughly 22% more likely to be cited, and adding direct quotations about 37% more likely, according to the Princeton-led study that first formalised GEO; applied together, such methods lifted visibility by up to 40%. Freshness compounds it: one 2025 analysis found about 65% of AI-bot visits target pages updated within the past year. The honest version of GEO and the useful version of content turn out to be the same thing.
| Signal | Finding | Source |
|---|---|---|
| Reddit citation share | Cited in ~40.1% of AI answers | Semrush, Jun 2025 |
| Wikipedia citation share | Cited in ~26.3% of AI answers | Semrush, Jun 2025 |
| Wikipedia for factual queries | ~47.9% of ChatGPT's top factual sources | Frase, 2026 |
| Concentration | ~15 domains ≈ 68% of AI citations | 5W Citation Index, 2026 |
| Overlap with Google | Only ~12% of AI citations match Google's top 10 | Ahrefs, Aug 2025 |
| Statistics in content | +22% citation likelihood | Aggarwal et al., KDD 2024 |
| Direct quotations | +37% citation likelihood | Aggarwal et al., KDD 2024 |
| Freshness | ~65% of AI-bot visits target content updated within a year | Wellows, 2025 |
Reading the weights
A skincare brand wants to be recommended by ChatGPT and Perplexity when someone asks for the best gentle cleanser. Which move best fits how these engines source answers?
B is correct. Answer engines lean on community and review platforms and on concrete, attributable facts. A and D are keyword stuffing and spam that get discounted or removed; C gets reverted and can backfire. Section 6.6.
6.7 Designing to be found
If visibility now depends on usefulness, then being found is a content-design problem, not a trick applied at the end. Three working principles follow from this chapter.
| Principle | What it means | What it asks the team to give up |
|---|---|---|
| Answer the question | Resolve the searcher's real intent, not just match the words | Keyword-first content written for the engine, not the person |
| Be platform-native | Produce the trust signals the local platform actually rewards | The assumption that one global SEO playbook fits everywhere |
| Earn the citation | State facts clearly enough to be lifted and attributed | Marketing copy that persuades but cannot be referenced |
The thread running through all three is that the fundamentals did not die; they got redistributed. Relevance, authority, structure, and speed used to win rankings. Now they decide whether a model trusts you enough to put you inside its answer. The work is the same; the reward moved.
The chapter argument, in one sentence
Which statement best summarises Chapter 6?
D. The chapter is one argument: meet intent with usefulness, respect platform-native logic, and write to be referenced, not just ranked. Fundamentals carry across; the reward moved from the click to the citation.
· Worked example · The RANKS rubric
Before you publish a page meant to be found, run it through RANKS. Five quick checks, scored 1 to 5, totalled out of 25. Below 15 means the page may rank for a keyword but will struggle to be trusted, lifted, or cited. The worked card audits a small specialty roaster's coffee-origin page, the same brand carried through the discovery chapter. The blank card is for one of your own.
RANKS: Relevance, Authority, Newness, Klarity, Speed. Score each, then total.
Now do one yourself. Pick a page you want to be found, your own project or a brand you study, and run it through RANKS. If you score below 15, write one sentence naming the single change that would lift the lowest letter by one. If your weakest letter is Authority, ask what concrete fact you could add that a model would be willing to trust and attribute.
Tick the score, write one line of justification, then total.
Test yourself before discussion
True or false. Answer first, then read the explanation. If you miss more than one, revisit the section noted before continuing to Chapter 7.
-
Search behaviour reflects active intent, which is why search is a higher-value surface than broadcast interruption.
When users search, they are already motivated. Good SEO meets that intent rather than forcing a message onto an uninterested audience. Section 6.1.
-
Repeating a keyword many times is the most reliable way to make a page rank and stay ranked.
Relevance means answering the query, not matching its words. Optimisation without substance is fragile. Section 6.2.
-
A Google-shaped SEO strategy copied directly into Korea often fails because Naver search follows its own platform-native logic.
Channels are cultural artefacts. Global platforms sit on top of local behaviour; they do not erase it. Section 6.3.
-
As answer engines synthesise responses and cite sources, being found shifts from ranking high to being trusted enough to be referenced.
Users increasingly receive synthesised answers rather than lists of links. The unit of visibility moves from the click to the citation. Section 6.4.
-
The rise of AI search means SEO fundamentals like structure, relevance, and authority no longer matter.
The fundamentals did not die; they got redistributed. They now decide whether a model trusts you enough to cite you. Sections 6.4 and 6.6.
· Reflection
Discussion questions
- Why does search reflect intent in a way that a social feed or a billboard does not? What does that change for the marketer?
- Of the four ranking signals, relevance, authority, structure, speed, which do brands most often neglect, and why?
- What does a Naver results page reward that a Google results page does not? How would a foreign brand adapt?
- Has an AI answer engine ever decided a purchase or a fact for you without you clicking any source? What made it feel reliable?
- What is the practical difference between writing a page to win a click and writing it to be cited?
- If a model can lift a single clear fact and attribute it, is that a marketing opportunity or a loss of control over the message?
- Where is the line between making content genuinely useful and engineering it to be cited? Is there one?
- Which search surface is the brand you study weakest in, and what would it cost to enter it well?
- As conversational search grows, what happens to the value of a brand's own website?
- Will answer engines eventually make traditional SEO irrelevant, or simply raise the bar for what counts as useful content? Defend your view.
· Chapter summary
Search matters because it reflects intent: a query is a declaration of active demand, and good SEO meets that demand rather than interrupting it. Engines decide what to surface using durable signals, relevance, authority, structure, and speed, and those signals still matter even as they become table stakes. Search is also cultural: platform-native environments like Naver in Korea follow a local logic that a Google-shaped strategy misses, because global platforms sit on top of local behaviour rather than replacing it.
The largest shift is the move from ten blue links to a single synthesised answer. AI answer engines interpret questions, summarise responses, and cite a small set of sources, so being found stops meaning ranking first and starts meaning being trusted enough to be referenced. This rewards content that is specific, structured, and willing to state facts plainly; marketing copy without substance gets ignored because there is nothing for a model to lift and attribute. The fundamentals did not die; they got redistributed from winning the click to earning the citation. The RANKS rubric, Relevance, Authority, Newness, Klarity, Speed, is one way to keep that judgement defensible before you publish.
· References used in this chapter
The full bibliography is on the References page.
- Clement, M. (2026). Digital marketing: An integrated, project-based approach (3rd ed.). Independent practitioner publication. Chapter 5, AI, Search, and the Decline of Traditional SEO Assumptions.
- Moz. (2025). The beginner's guide to SEO. moz.com
- Naver. (2025). Advertising and business resources. business.naver.com
- Perplexity. (2025). How Perplexity sources answers. perplexity.ai
- OpenAI. (2024). Introducing ChatGPT Search. openai.com
- Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative Engine Optimization. Proceedings of the 30th ACM SIGKDD Conference (KDD 2024). arxiv.org/abs/2311.09735
- Semrush. (2025). The most-cited domains in AI: a 3-month study. semrush.com
- Peec AI. (2026). Top domains cited by AI search: analysis based on 30M sources. peec.ai
- Ahrefs. (2025). AI citations vs Google rankings. ahrefs.com
- Statista. (2025). Internet usage in South Korea. statista.com
- Think with Google. Moments of intent. thinkwithgoogle.com