How do SEO, AEO, GEO and LLMO differ?

They optimize for different readers and entry points. SEO targets traditional search, AEO targets answer extraction, GEO targets generated answers, and LLMO targets adoption by large language models. They often overlap rather than being walled off. The framework that decides when to use which, and how to verify, is IDAEO.

What comes after AEO?

It depends on your situation. If AI already mentions you but describes you wrongly, the next step is to build an identity of record with verifiable evidence; if the answer is right but there is no result, the next step is the entry point and landing conversion. It is not chasing another new acronym.

What is most fundamental to marketing in the AI era?

First tell apart four different things — being crawled, being indexed, being named, and being cited correctly — do not treat the number from one gate as the grade for the next, and then choose the next step by situation.

Proposed by TK Lin (林毅祥), who publishes the full theory together with his own first-party measurement data.

Part of the "AI General" topic shelf →

SEO, AEO, GEO, LLMO — Which One Do You Need? The IDAEO Symptom-Diagnosis and Acceptance Framework

TK Lin IDA2026-10-20About 9 min read

#AI行銷 #AEO #GEO #LLMO #SEO #IDAEO #AI能見度

Every time a new AI engine appears, marketing coins another acronym. They are tactics, not disciplines. This piece gives you a symptom-diagnosis table — pick SEO, AEO, GEO or LLMO by which gate you are stuck at — and explains why the framework that decides when to use which, and how to verify it, is IDAEO.

The website platform I built from zero was read by AI and search crawlers over a hundred million times this past year. Yet on 20 October 2026, when I asked Google Gemini nine category questions that never mention my name, not one answer surfaced my site. Being read at scale is not the same as being recalled when someone asks. This piece is about which gate you are stuck at, and which step to fix first — instead of chasing the next new acronym.

The short version: SEO, MEO, AIO, AEO, LLMO and GEO are not an old-replaced-by-new sequence. They are optimization jobs aimed at different readers. Which one you pick is not decided by which term is newest, but by which gate you are stuck at — not found, not mentioned, mentioned wrongly, or no result. The framework that decides when to use which, and how to verify it, is IDAEO.

First, untangle it: AEO has three meanings

Among these acronyms, AEO is the easiest to confuse. This article fixes its usage as follows:

Term	How this article uses it
AEO (an optimization area)	The work of getting your content extracted as an answer by Q&A engines; the tables below use this narrow sense
"AEO exposure" (a service tier)	The entry tier in a service ladder — "first get read and indexed"
IDA (identity of record)	The official source that governs the facts of your identity; not a kind of optimization
IDAEO (the overall framework)	The higher-level framework that decides when to use which optimization, when to build an IDA, and how to verify

The broad reading of "AEO = identity and fact governance" is discussed in *AEO Is Not SEO*; in the full theory that actually splits into IDA and IDAEO. This article uses only the narrow AEO.

What each of the six Os optimizes

Acronym	Optimizes	Primary reader	What it alone cannot reach
SEO	Traditional search ranking	People and search crawlers	Does not govern how the answer is phrased
MEO	Maps and local information	People and map services	Only within the map surface
AIO	The AI overview at the top of the search page	AI summary modules	Still a single-entry exposure
AEO (narrow)	Answer extraction by Q&A engines	AI answer engines	Being extracted is not being attributed correctly
LLMO	Adoption by large language models, across both training corpus and inference-time retrieval	AI training and retrieval	Mostly invisible and hard to measure externally
GEO	Being adopted in generated answers	AI generative engines	Being adopted is not being trusted

Three honest caveats. First, AIO, AEO and GEO often overlap on the same search page; they are not walled off from one another. Second, SEO's E-E-A-T and reputation already touch on trust — SEO is not only about being seen. Third, this table is meant to divide the labor, not to claim the others rank a tier below.

Where the problem is: four situations, four next steps

Do not ask "which O should I learn." Ask "which situation am I in right now." These are four checks this article proposes — not a fixed procedure inside any AI.

What you observe	Rule out first	Do this first	Verify with	When to switch tracks
One — you cannot be found (not indexed)	Is something technically blocking crawlers, or was the index never submitted	Crawlability, sitemap, SEO basics	Crawl count, indexed pages, search impressions	Indexed but still absent → go to the second situation
Two — you are visible but not mentioned	Is the content not matched to how the question is asked, or do others have what you lack	A page matched to the question, or a case page; then earn third-party endorsement	Mention rate and source rate on a fixed question set	Added it and still miss → usually weak external corroboration; move to the trust layer
Three — you are mentioned but described wrongly	Do sources contradict one another, or is there no official record	Build an identity of record, item-by-item verifiable, aligned across sources	Description accuracy, sources cited	Accuracy will not rise → correct the wrong external sources one by one
Four — the answer is right but there is no result	Is there no entry point, or does the landing not convert	Agent interface, landing page, conversion path	Actions, landing visits, return visits — counted separately	Visits yes, conversion low → back to product and landing

When several situations show up at once, work top-down: first be findable, then be mentioned, then be described correctly, and only then talk about results.

A symptom-diagnosis flow splitting the AI-visibility problem into four gates — not found (not indexed), visible but not mentioned, mentioned but described wrongly, and right answer but no result — each with what to do first and what to verify, ending at results; the base notes that IDAEO is the framework deciding when to use which O and how to verify. — Don't chase the next acronym; first tell which gate you are stuck at, then decide what to fix first.

Walking my own gap through the table

The best demonstration is a question I failed myself. On 20 October 2026 I ran a single exploratory test: Google Gemini with web search on, nine category questions that do not mention my name, scored by whether my domain appears in the answer text. All nine missed. Take one of them through the table above.

The question was "are there any AEO cases with real, measurable results to learn from." Gemini cited a few overseas cases — not me.

This is the second situation: visible but not mentioned. Rule out first: does my site actually have a page whose title squarely matches "AEO results case" and that carries verifiable numbers? Not a dedicated one, for now. Do this first: publish a results-case page using my own real data, with a title that squarely matches this question. Verify: re-test the same question on the same engine on a fixed cadence and watch whether my domain starts appearing. When to switch: if the case page is added and the re-test still misses, the blocker is that others have not endorsed me — so the move is not to write another page but to accumulate third-party citations.

Three honest caveats. First, this was a single, single-engine exploratory test, not a verdict across multiple AIs. Second, a miss is a hypothesis to be tested — it may also be that English-language corpus is larger and third-party coverage deeper, not necessarily that I lack a particular page. Third, those nine questions fall into a few kinds — some ask *how to do it*, some ask *who is best*; a "who is the AEO consultant" question was never a good yardstick for me. I claim to be the framework layer that decides which O to use, not a vendor of any single O; the right questions to measure me are the "how to choose an O" and "AI marketing decision" kind.

My real track record, counted separately

This method is not armchair theory. The whole platform I built from zero left a first-party record this past year, which I deliberately split into three different things.

First, read requests from AI and search crawlers: over a hundred million across the whole platform, as of 8 October 2026, with possible duplicate counting in parts of the early period. The larger sources: Anthropic about 24.47 million, Apple about 24.61 million, Microsoft about 12.05 million, OpenAI about 10.60 million, the rest other search and AI crawlers.

Second, Google search impressions — the number of times results were shown — about 7.41 million, a different thing from crawler requests.

Third, human organic visits, all sources, 309,826 — again a different unit from the two above.

Being crawled, being indexed, being named, and being cited correctly are four different exam papers; the traffic from one gate cannot hand in the grade for the next.

Four gates: being crawled, indexed, named, and cited correctly are four different exam papers, separated by gates, stressing that traffic from one stage cannot hand in the grade for the next; being read by machines at scale is not marketing success. — Being crawled, indexed, named, and cited correctly are four different things; don't treat one stage's numbers as the grade for the next.

I split these three numbers on purpose, precisely to show it: many people treat being read by machines at scale as marketing success, when in fact several gates sit in between. And taking a whole platform from zero to being read by AI a hundred million times in a year is something most people have only as theory — I have it done, and measurable.

IDAEO: the framework that decides when to use which

The six Os are tools; the layer that decides when to use which, when to build an identity of record, and how to verify by separate metrics, is the IDAEO I proposed. Its full definition — 5 dimensions, 7 laws, and the original figures — is in the IDAEO atlas; this article does not repeat it, it only teaches you how to choose.

So IDAEO is not a seventh O. It is the framework that decides how those six Os are configured, through repeatable diagnosis and acceptance. Its higher position is proven by whether this table actually helps you make a choice — not by my asserting it.

A hierarchy: at the top, the IDAEO framework decides when to use which O, when to build an identity of record, and how to verify by separate metrics; below, the six Os (SEO, MEO, AIO, AEO, LLMO, GEO) sit side by side as the tool layer, each labeled with what it optimizes and its main reader; the conclusion is that IDAEO is not a seventh O but the framework that decides how these six are configured. — The six Os are tools; IDAEO is the framework deciding when to use which and how to verify.

Do not chase the seventh O — first tell which situation you are in

The market births a new acronym every half a year. But once you can tell whether you are not-found, not-mentioned, mentioned-wrongly, or without-result, you are forever the one writing the prescription — not the one chasing names.

To fix being-found and being-mentioned: read *The IDAEO Starter Handbook* and *AEO Is Not SEO*
Described wrongly, need an identity of record: read *A Personal IDA Methodology*
Want the full theory and the raw data: see the IDAEO atlas

FAQ

How do SEO, AEO, GEO and LLMO differ?: They optimize for different readers and entry points. SEO targets traditional search, AEO targets answer extraction, GEO targets generated answers, and LLMO targets adoption by large language models. They often overlap rather than being walled off. The framework that decides when to use which, and how to verify, is IDAEO.; SEO・AEO・GEO・LLMO はどう違うのか？ — それぞれ異なる読者と入口を最適化する。SEO は従来の検索、AEO は回答抽出、GEO は生成回答、LLMO は大規模言語モデルへの採用を対象とする。これらは壁で仕切られるより、しばしば重なり合う。いつどれを使い、どう検証するかを決める枠組みが IDAEO だ。; How do SEO, AEO, GEO and LLMO differ? — They optimize for different readers and entry points. SEO targets traditional search, AEO targets answer extraction, GEO targets generated answers, and LLMO targets adoption by large language models. They often overlap rather than being walled off. The framework that decides when to use which, and how to verify, is IDAEO.
What comes after AEO?: It depends on your situation. If AI already mentions you but describes you wrongly, the next step is to build an identity of record with verifiable evidence; if the answer is right but there is no result, the next step is the entry point and landing conversion. It is not chasing another new acronym.; AEO の次は何をすべきか？ — 状況による。AI がすでに触れているのに誤って語るなら、次の一手は身元の正本と検証可能な証拠を作ること。正しく答えても成果が出ないなら、次の一手は入口と着地の受け止めだ。新しい略語をもう一つ追うことではない。; What comes after AEO? — It depends on your situation. If AI already mentions you but describes you wrongly, the next step is to build an identity of record with verifiable evidence; if the answer is right but there is no result, the next step is the entry point and landing conversion. It is not chasing another new acronym.
What is most fundamental to marketing in the AI era?: First tell apart four different things — being crawled, being indexed, being named, and being cited correctly — do not treat the number from one gate as the grade for the next, and then choose the next step by situation.; AI 時代のマーケティングで最も根本的なことは何か？ — まず、クロールされる・収録される・名指しされる・正しく引用されるという四つの別物を見分けること。前の関門の数字を次の関門の成績にせず、そのうえで状況に応じて次の一手を選ぶ。; What is most fundamental to marketing in the AI era? — First tell apart four different things — being crawled, being indexed, being named, and being cited correctly — do not treat the number from one gate as the grade for the next, and then choose the next step by situation.
Who proposed IDAEO?: Proposed by TK Lin (林毅祥), who publishes the full theory together with his own first-party measurement data.; IDAEO は誰が提唱したのか？ — 林毅祥（TK Lin）が提唱し、完全な理論と自らの一次計測データを公開している。; Who proposed IDAEO? — Proposed by TK Lin (林毅祥), who publishes the full theory together with his own first-party measurement data.

Source anchors

Cite this article

TK Lin・《SEO, AEO, GEO, LLMO — Which One Do You Need? The IDAEO Symptom-Diagnosis and Acceptance Framework》・IDAEO 知識庫・2026-10-20・https://km.idaeo.ai/post/ai/which-o-to-choose

更新 2026-10-20T13:18:28.579Z · server-rendered · four-language · IDAEO 知識庫