AI search, GEO, AEO and LLMO: a field guide to the terms

AI-mediated discovery is developing faster than its vocabulary. The result is a cluster of terms that are often used as synonyms even when they describe different layers of the same problem.

This field guide is York Studio's working map. It will change when the products, methods, and evidence change.

The short version

AI search is the environment: people use an AI system to discover, compare, explain, or decide something.

AI visibility is the measurable outcome: whether, where, how, and with what evidence an entity appears in those answers.

GEO, AEO, LLMO, and AI SEO are overlapping practice labels. They describe attempts to make information easier for search engines and language-model systems to find, interpret, retrieve, cite, and recommend.

AI search

AI search is the broadest useful term. It includes conversational systems with web search, grounded answer engines, AI summaries inside conventional search, shopping or local answer surfaces, and emerging agent-led discovery.

The important distinction is not whether the interface looks like a search box. It is whether an AI system participates in selecting, synthesising, or recommending the information a person receives. OpenAI, Anthropic, and Google all document forms of web-connected or search-grounded model use, but each surface has different retrieval, citation, location, personalisation, and product behaviour.

AI visibility

AI visibility describes what can be observed about an entity inside an AI answer. That can include mention rate, recommendation rate, citation rate, the descriptions attached to a brand, competitor co-mentions, factual accuracy, source influence, and variation over time.

Visibility is not one universal rank. A result belongs to a prompt, a surface, a market, a moment, a method, and usually a sample. Removing that context creates a clean number at the cost of an honest one.

GEO: generative engine optimisation

GEO usually refers to work intended to improve representation in generative or answer-engine outputs. Its useful focus is the complete evidence environment: first-party pages, structured facts, third-party sources, product feeds, reviews, citations, and entity consistency.

The term is imperfect because it can imply a stable "engine" that can be optimised like a conventional ranking system. In practice, the observable output may depend on retrieval, generation, product rules, model versions, and sampling variation.

AEO: answer engine optimisation

AEO predates much of the current LLM discussion. It grew from featured snippets, voice answers, knowledge panels, and structured answers. Today it is often used more broadly for being selected or cited in AI-generated answers.

Its strength is the emphasis on answering a question clearly. Its weakness is that it can narrow the work to page formatting when the answer may depend on an ecosystem of external sources and product-specific retrieval systems.

LLMO: large language model optimisation

LLMO focuses specifically on language-model systems. It is useful when discussing how models interpret entities, claims, context, and source material. It becomes misleading when it treats a model as the only component involved: many commercial experiences add search, ranking, grounding, safety, personalisation, or interface logic around the model.

AI SEO

AI SEO is the most accessible bridge from conventional search practice. It signals that technical accessibility, content quality, authority, and structured information still matter while the discovery interface changes.

The risk is assuming that familiar ranking concepts transfer unchanged. AI answers can mention a brand without linking, cite a page without recommending its owner, or make a recommendation assembled from several sources.

How York Studio uses the terms

York Studio uses AI search and discovery for the whole field and AI visibility for observed representation inside it. GEO, AEO, LLMO, and AI SEO remain useful subtopics, search terms, and practice labels. None is treated as a promise that a particular intervention will cause a particular answer.

That hierarchy keeps the subject larger than today's acronym—and leaves room for agents, commerce, local discovery, and new interfaces that have not yet acquired a name.

Required reading

Limitations

• The vocabulary is not standardised and different vendors use the same term differently.
• Product behaviour changes faster than durable terminology, so this guide requires periodic revision.
• This is a conceptual field map, not a comparative performance study.

Primary references

Sources

1. OpenAI: web search tool
2. Anthropic: web search tool
3. Google: grounding with Google Search