This is the first entry in a new reference series on this blog. Everything else here is written as an explainer or a guide. This is different on purpose: short, direct definitions for the terms that come up constantly in AEO and AI search discussions, organized for quick lookup rather than a single narrative read. Where a term has a full dedicated guide elsewhere on this blog, it's linked below the definition.
Core Concepts
AEO (Answer Engine Optimization). The practice of structuring content and technical infrastructure so AI systems can find, understand, and cite it when generating an answer, as distinct from optimizing purely for a traditional ranked list of links. See the full AEO guide.
GEO (Generative Engine Optimization). A closely related term, often used interchangeably with AEO, specifically emphasizing optimization for generative AI systems that synthesize an answer rather than simply listing sources. See the GEO versus AEO versus SEO guide for how the terms actually differ in practice.
Citation. An instance of an AI system referencing or attributing a specific claim to a specific source in a generated response, whether displayed as a visible link, an inline reference, or simply named in the response text.
Zero-click search. A search interaction that ends without the user clicking through to any external website, because an AI-generated summary or direct answer satisfied the query within the results interface itself. See the zero-click economy guide.
Retrieval and Model Mechanics
RAG (Retrieval-Augmented Generation). The architecture where an AI system retrieves external content at query time and uses it to ground a generated response, rather than answering purely from what the model learned during training. See the RAG pipeline guide.
Parametric memory. What a language model "knows" purely from its training data, with no live retrieval involved — the opposite of retrieval-augmented generation. See the parametric versus retrieval memory guide.
Query fan-out. The technique where an AI system decomposes a single user query into multiple related sub-queries, retrieving and synthesizing across all of them before producing one response. See the query fan-out guide.
Vector embedding. A numerical representation of text's meaning, used by retrieval systems to measure how semantically similar a piece of content is to a query, independent of exact keyword overlap. See the vector embeddings guide.
Cosine similarity. The specific mathematical measure most retrieval systems use to score how closely a content embedding matches a query embedding, expressed as the angle between the two vectors.
Chunking. The process of splitting a page's content into smaller segments before embedding and retrieval, so a system can match and cite a specific passage rather than an entire page.
Attention mechanism. The part of a language model's architecture that weighs which parts of an input are most relevant when generating a response, which has a documented tendency to weight information at the beginning and end of a long input more heavily than the middle. See the attention mechanism and position bias guide.
Content Structure Terms
BLUF (Bottom Line Up Front). A writing structure that states the direct answer or conclusion in the first sentence, before supporting detail or context, optimized for how AI extraction systems tend to weight early content most heavily. See the BLUF writing guide.
Non-commodity content. Content that provides a genuine, unique point of view, original data, or documented first-hand experience, as distinct from generic content a language model could plausibly synthesize on its own without needing to cite an external source. See the non-commodity content guide.
Content decay. The gradual reduction in an AI system's confidence in a piece of content's accuracy or relevance as it ages without updates, distinct from traditional SEO decay. See the content decay guide.
Citation half-life. The typical time period after which roughly half of a piece of content's citation activity on a given engine has faded, used as a practical benchmark for how often that specific engine expects freshness updates.
Technical and Schema Terms
Schema markup (JSON-LD). Structured data embedded in a page's code that explicitly labels its content type and key facts for machine readability, distinct from the visible text a human reader sees. See the schema markup master guide.
llms.txt. A proposed text file, similar in spirit to robots.txt, intended to give AI systems a curated, machine-readable summary of a site's key content and structure. See the llms.txt mechanics guide for what it actually does and doesn't influence.
Content-Signal directive. A robots.txt extension that lets a site declare separate permissions for search indexing, AI input use, and AI training use, rather than a single blanket allow or disallow.
Canonical tag. An HTML element declaring which version of a page should be treated as the authoritative source when duplicate or near-duplicate content exists across multiple URLs. See the canonical tags guide.
SSR (Server-Side Rendering) versus CSR (Client-Side Rendering). Two approaches to building a webpage, where SSR delivers fully-formed HTML content immediately and CSR requires JavaScript execution to populate the page — a distinction that matters enormously for AI crawlers, most of which do not execute JavaScript. See the SSR versus CSR guide.
Trust and Authority Terms
E-E-A-T. Experience, Expertise, Authoritativeness, and Trustworthiness — the framework describing the credibility signals search and AI systems weigh when evaluating a source. See the E-E-A-T for AEO guide.
Entity disambiguation. The process by which an AI system resolves which specific real-world entity a name or reference refers to, particularly important when a brand name is inconsistent or ambiguous across the web. See the entity disambiguation guide.
Topical authority. The degree to which a domain demonstrates comprehensive, interconnected expertise across a specific subject area, as distinct from having one or two strong individual pages with no broader supporting structure. See the topical authority guide.
Unlinked mention. An instance of a brand or entity being named on a third-party page with no accompanying hyperlink, which independent research has found correlates more strongly with AI citation likelihood than a traditional backlink does. See the brand mentions guide.
AI Share of Voice. A measurement of how frequently a brand is cited or mentioned across AI-generated answers for a defined set of relevant queries, relative to competitors answering the same queries. See the AI share of voice guide.
Brand hallucination. An instance of an AI system generating an inaccurate claim about a specific brand, product, or company, whether from outdated training data or a retrieval error. See the brand hallucination guide.
How This Glossary Will Grow
This first entry covers 25 foundational terms. As the vocabulary around AI search continues to evolve — and it has changed noticeably even within the timeframe this blog has been publishing — this glossary will be revisited and extended rather than treated as a one-time, static reference. If there's a term you've encountered that isn't defined clearly anywhere, including here, that's useful feedback for what to add next.
This glossary intentionally keeps definitions short and links out to a full guide wherever one exists on this blog. If you're looking for the mechanics behind a term rather than just its definition, follow the link — that's where the actual depth lives.
Frequently Asked Questions About This Glossary
Will this glossary be updated as terminology changes?
Yes. AI search terminology has shifted meaningfully even over a relatively short period, and this reference is intended to be revisited rather than left static once published.
Is there a difference between AEO and GEO worth remembering beyond this glossary entry?
The two terms are used close to interchangeably in most practitioner conversation, and the dedicated guide linked above covers the more precise distinctions some practitioners draw between them, along with where they genuinely overlap with traditional SEO.
Can I suggest a term for a future glossary update?
Yes — this is exactly the kind of reference content that benefits from real practitioner input about which terms are actually causing confusion in the field, rather than an editorial team's guess at what's missing.