The terms that come up constantly in AI search, defined plainly. Where the industry uses a word loosely, we say so rather than pretending there's a settled definition.
Sixteen terms that do most of the work in AI search conversations. Each links to a page covering it properly.
Optimizing so generative AI systems cite your brand as a source when composing an answer. Rests on retrievability, entity resolution, corroboration, and extractable structure.
Read more →Structuring content so an AI system can extract a direct, self-contained answer from it. Narrower and more structural than GEO, and usually faster to move.
Read more →The umbrella term for optimizing toward language-model retrieval and memory. Sometimes written LLMO.
Read more →AI-generated answers appearing above organic results in Google Search. The highest zero-click risk surface currently.
Read more →OpenAI's web crawler, controllable via robots.txt. Separate from OAI-SearchBot, which serves search and citation rather than training.
Read more →A robots.txt directive controlling whether your content is eligible for Gemini grounding. Separate from Googlebot and doesn't affect rankings.
Read more →A company, product, person, or concept a system can identify and attach attributes to — as opposed to an ambiguous text string.
Read more →A schema property linking your entity to verified external profiles, so independent sources confirm one identity rather than several.
Read more →The proportion of a benchmarked prompt set where your brand is named in the answer, tracked per platform over time.
Read more →Whether the content a crawler receives matches what a browser renders. The most common silent AI visibility failure.
Read more →A search or query answered completely in the interface, so the user never visits a source. Rising sharply on informational queries.
Read more →A proposed file describing site content for language models. Adoption is genuinely uncertain; cheap to implement, not a strategy.
Read more →Multiple credible sources stating the same fact. Models weight this heavily — a claim only on your own site carries far less weight.
Read more →Whether a passage stands alone well enough to be lifted into an answer without losing accuracy or context.
Read more →A defined map of your core topic and every concept within it, used to assign canonical pages and prevent cannibalisation.
Read more →Multiple pages on your own site competing for the same concept, splitting signals and confusing which page is definitive.
Read more →GEO, AEO, and LLM SEO get used interchangeably across the industry, and a lot of the apparent disagreement is people defining them differently rather than actually disagreeing.
We use GEO for citation within generated answers, AEO for being the extracted direct answer, and LLM SEO as the umbrella. That's a convention, not a standard — and in practice roughly 80% of the underlying work is shared regardless of what you call it.
Be sceptical of anyone selling these as three separate products with three separate retainers.
GEO generally means being cited as a source inside a generated answer. AEO means being the extracted direct answer to a specific question. LLM SEO is usually the umbrella term for optimizing toward language models overall.
The distinctions are real but narrower than the marketing suggests. Around 80% of the underlying work — crawler access, entity clarity, corroboration, extractable structure — is shared across all three.
Partly, and anyone claiming it's entirely new is overselling. The foundation is shared: crawlable, fast, credible, well-structured sites.
What's genuinely different is the failure mode. In classic search, weak signals mean ranking lower. In AI search, a blocked crawler or an unresolved entity means total absence, with no degraded position to fall back on.
The proportion of a fixed, benchmarked prompt set where your brand is named in the answer, tracked per platform.
It's noisier than rank tracking because model outputs vary between runs. Reported properly it includes variance rather than presenting a single run as fact.
No, and an agency that requires you to is making things harder than necessary. The underlying questions are simple: can systems read your site, do they know who you are, can they quote you, and does anyone else vouch for you.
Everything in this glossary is a more precise way of asking one of those four.
We'll benchmark your brand against the questions your category actually gets asked, and show you who's being cited instead of you.
No commitment · 45 minutes · Immediate value