AEO & GEO Glossary
Sixty definitions covering answer engine optimization, generative engine optimization, share of voice and the proprietary metrics SkuLift uses to measure brand visibility inside AI answers.
What is this glossary?
A reference glossary of sixty AEO and GEO terms, from answer engines and structured data to SkuLift's own share-of-voice metrics, written to be cited by AI search engines.
What is AEO, GEO and answer-engine visibility?
The two disciplines this glossary revolves around, and why the distinction matters.
Answer Engine Optimization, or AEO, is the practice of making a brand's existing content easier to cite when an AI engine answers a question. It works on what already exists: product pages, documentation, comparison tables and help articles are restructured so a model can extract a clean, attributable answer. The goal is not a blue link but a sentence the engine repeats, with your brand named as the source.
Generative Engine Optimization, or GEO, is the complementary discipline. Instead of reshaping existing pages, GEO produces new content designed from the start to be referenced by conversational AI engines. A definitive glossary, a methodology page, a head-to-head comparison or an authoritative FAQ are all GEO assets: they exist to occupy a question that has no good answer yet, so the model has something to ground on and a clear entity to attribute.
Both disciplines target the same surfaces. An answer engine is an AI system that responds to a query directly rather than returning a list of links. A generative engine produces fresh prose, summaries or recommendations on demand. ChatGPT, Claude, Gemini and Perplexity are the engines that matter for most brands, and each one decides independently which sources to cite and how prominently to name them.
Visibility inside those answers is measurable, and that is the thread running through this glossary. A brand that is never mentioned is invisible regardless of how good its website is. A brand that is mentioned, cited and quoted in a favourable position is winning the answer. The terms below give you the vocabulary to describe exactly where a brand sits on that spectrum, engine by engine and query by query.
A useful way to hold the distinction is by asset and by intent. AEO assets are pages you already publish and now make extractable; GEO assets are pages you create because a question is unowned. The same metric set measures both, which is why this glossary treats AEO and GEO as two routes to one outcome: being the named, quoted source when an engine answers a buyer's question.
If you are new to the field, start here and treat the rest of the glossary as the supporting detail. Everything downstream, every metric and protocol, exists to answer one question for a specific brand on a specific engine: are you in the answer, where, and how favourably? AEO and GEO are simply the two ways to change that answer in your favour.
How to read this glossary
Every card follows the same shape, so you can scan or go deep without surprises.
Each term opens with a canonical definition of forty words or fewer. This is deliberate: short, self-contained definitions are the unit an AI engine extracts and repeats. If you read nothing else, the opening line is the answer. Below it sits an extended definition that adds context and explains why the term matters for AI search specifically, not just for classic SEO.
After the definition comes a concrete example, usually drawn from a real share-of-voice campaign, so the abstract idea has something to attach to. Then a short list of related terms links across the glossary. This internal mesh is intentional: the leaders in this field hold their position partly by densely cross-linking their lexicon, which signals to engines that the vocabulary is owned and coherent rather than scattered.
Terms fall into three groups. Foundational AEO and GEO concepts come first. Then the protocols and techniques that make agentic commerce work, from retrieval-augmented generation and structured data to the Model Context Protocol. Finally the market and method vocabulary: AI Overviews, zero-click search, entity SEO, web grounding versus parametric memory, and the query types that segment a measurement run.
Proprietary SkuLift terms are marked by their precision. Metrics such as Position-Weighted Count, Word Count Share, Citation Authority Score and the four-level SOV pyramid are defined exactly as they are implemented in the platform. We do not paraphrase them loosely, because a glossary that contradicts the product it documents teaches an engine the wrong answer. Where a definition is proprietary, it is the definition.
You can read the glossary two ways. Skim the A-to-Z grid for the one-line definition of a term you half-remember, or open a card and follow its related links to walk the field as a connected graph. Both are valid; the second is how engines crawl it, hopping from a definition to its neighbours and inferring that the vocabulary belongs to a single coherent source.
Finally, the glossary is bilingual and kept at strict parity. The English slug is canonical and the French page is served under the same URL with a locale prefix, so a term has one identity across languages. That parity is enforced automatically, which means a definition can never exist in one language and silently go missing in the other.
SkuLift proprietary terms
The metrics and product concepts SkuLift coined, defined once, here.
SkuLift forged a set of terms to make AI visibility rigorous rather than anecdotal. Share of Voice is the core measurement: multi-engine, multi-level, and decomposed into a four-level pyramid that separates raw presence from genuine quality. The pyramid runs from Presence, a binary mention rate, through Volume, through Quality, up to weak signals such as variance and query drift. Each level answers a different question about how a brand actually appears.
The quality metrics are where the rigour lives. Position-Weighted Count applies an exponential position decay to brand sentences, rewarding mentions that appear early in an answer where readers and engines weight them most. Word Count Share measures the proportion of an answer that talks about the brand at the sentence level. Citation Authority Score grades the authority of the sources the brand itself owns, not every URL the engine happens to cite. First-Mention Position records, as a fraction of the answer, how early a brand first appears.
Beyond metrics, SkuLift's product vocabulary describes how measurement turns into action. A lift is a packaged automation that delivers a measurable change to a brand surface, versioned and replayable with before-and-after KPIs. A studio is the table-shaped data space that feeds it. The brand kit is the single source of truth for identity, and the knowledge base is the per-workspace document store that grounds recommendations. Tying them together is the closed loop: measure, understand, recommend, gate, execute, re-measure, remember.
Governance is explicit in this vocabulary too. The human gate is the approval step an owner must clear before anything is written to an external surface, enforced at three levels of the system. The agentic commerce protocols, ACP, AP2 and MCP, name the standards SkuLift speaks to ChatGPT, Gemini and Claude respectively. Defining these terms publicly is itself a GEO move: SkuLift coined the category, so the glossary makes SkuLift the citable source for it.
One more reason precision matters here: these metrics are reproducible. N-sampling fixes five samples per query per engine so averages and variance are stable rather than the product of a single lucky run. A/B/C classification then labels each measurement event by how consistently a brand appears across those samples, separating durable presence from noise. Reproducibility is what lets a number become a baseline you can move.
Taken together, these terms form a measurement language. Without it, AI visibility is a collection of anecdotes: someone asked ChatGPT and the brand came up, or did not. With it, visibility becomes a set of numbers you can track weekly, attribute to specific actions, and improve deliberately through the closed loop. That shift, from anecdote to instrument, is the reason the proprietary vocabulary exists.
Glossary FAQ
The questions people ask before they read the cards.
A few recurring questions are worth answering up front, because they shape how the rest of the glossary reads. The answers below are short on purpose; each links into the relevant term cards for the full treatment.
If your question is not answered here, the individual term cards go deeper, and the methodology and share-of-voice pillar pages explain how the metrics fit together end to end. The glossary is the index; those pages are the long form.
Browse the glossary A to Z
Sixty terms, alphabetised. Jump to a letter, then open any card for the full definition, an example, and its linked terms. Each card links to a dedicated page marked up as a DefinedTerm so search engines can index it on its own.
A
- A/B/C Classification
A/B/C classification is a measurement-event scheme by stability across the N samples: A events are stable across all samples, B events appear…
- ACP (Agentic Commerce Protocol)
ACP (Agentic Commerce Protocol) is the protocol ChatGPT (OpenAI) uses for agent-driven commerce flows. SkuLift implements ACP under src/app/api/acp/ to participate in…
- AEO (Answer Engine Optimization)
Answer Engine Optimization (AEO) is the practice of making a brand’s existing content easier for AI engines to cite when they answer…
- Agentic Commerce
Agentic commerce is commerce driven by AI agents: software that discovers products, compares options and completes purchases on a user’s behalf, through…
- Agentic Commerce Platform
An agentic commerce platform is the orchestration layer that makes a brand discoverable, citable and consistent across AI agents and engines. SkuLift…
- AI Overviews
AI Overviews are the AI-generated answers Google places at the top of Search, summarising the web and citing a handful of sources…
- AI Search
AI search is search where an engine returns a synthesised, often cited answer instead of a ranked list of links — spanning…
- AI Shopping Agent
An AI shopping agent is an AI that researches, compares and can complete purchases on a buyer's behalf — increasingly transacting through…
- Answer Engine
An answer engine is an AI system that responds to a query with a direct, synthesised answer — and increasingly its source…
- Answer Engine Optimization
Answer Engine Optimization is the full name of AEO: the practice of making a brand’s existing content easier for AI engines to…
- Answer-First
Answer-first is structuring content so the direct, self-contained answer comes first, before context or background. It is the editorial move that makes…
- AP2 (Agent Protocol)
AP2 (Agent Protocol) is the protocol used in Google’s Gemini agentic stack. SkuLift supports AP2 as future-proofing: current usage is narrow, but…
- Authority Signal
An authority signal is any piece of evidence an AI engine uses to judge whether a source is trustworthy enough to cite…
B
- BLUF (Bottom Line Up Front)
BLUF (Bottom Line Up Front) is a writing discipline that puts the conclusion or key message in the opening line, before any…
- Brand Entity
A brand entity is your brand represented as a recognised thing in a knowledge graph — with a stable identity, defined attributes…
- Brand Kit
The brand kit is SkuLift’s source of truth for a brand’s identity — logo, palette, fonts, voice anchors, ICP and positioning. It…
C
- CAS (Citation Authority Score)
Citation Authority Score (CAS) is the L3 quality component scoring the mean authority of the client-owned citations only — the brand’s and…
- Citation
A citation is a reference to a source by an AI engine — when it names, links to, or draws its answer…
- Citation Tracker
A citation tracker is the component that detects and attributes the brand and product citations inside an AI answer — finding where…
- Closed Loop
The closed loop is SkuLift’s product metaphor: a continuous cycle that runs measure, understand, recommend, gate, execute, re-measure and remember. Seven verbs…
- Conversational Search
Conversational search is searching by asking questions in natural language — often over several turns — and receiving a single synthesised answer…
- CPCS (Cross-Platform Consistency Score)
The Cross-Platform Consistency Score (CPCS) is a cross-engine consistency score: it measures how stable a brand’s mentions and citations are from one…
E
- E-E-A-T
E-E-A-T is Google's quality framework — Experience, Expertise, Authoritativeness and Trustworthiness. It describes the qualities that make content credible enough for engines…
- Entity SEO
Entity SEO is search optimisation organised around entities and their relationships in a knowledge graph, rather than around individual keywords — making…
F
- FAQ Schema
FAQ schema is schema.org FAQPage markup that labels question-and-answer pairs in a page so AI engines and search can parse them explicitly.…
- FMP (First Mention Position)
First Mention Position (FMP) is the first brand mention as a fraction of the response [0,1] — 0 is the very start,…
G
- Generative Engine
A generative engine is an AI engine that produces an original, composed answer by drawing on a large language model and, often,…
- Generative Engine Optimization
Generative Engine Optimization is the full name of GEO: producing new content designed to be referenced by conversational AI engines. It is…
- GEO (Generative Engine Optimization)
Generative Engine Optimization (GEO) is the practice of producing new content designed to be referenced by conversational AI engines. It is distinct…
H
- Hallucination
A hallucination is an AI answer that is factually false yet expressed fluently and confidently, as if true. It happens when a…
- Human Gate
The human gate is the explicit owner-approval step in the closed loop: a recommendation must be approved at the human gate before…
J
- JSON-LD
JSON-LD is a JSON-based syntax for embedding linked structured data in a page. It is the dominant way to express schema.org types…
K
- Knowledge Base
The knowledge base is SkuLift’s per-workspace documentary store: documents are uploaded, vectorised and indexed. It feeds KB-derived pillar recommendations and the AEO-GEO…
L
- Lift
A lift is a packaged automation that ships a measurable change to a brand surface — a published article, a Shopify update,…
- LLM (Large Language Model)
A Large Language Model (LLM) is a neural network trained on vast amounts of text to predict and generate language. LLMs power…
- LLM Visibility
LLM visibility measures how present and cited a brand is in the answers large language models give. SkuLift operationalises it through share…
M
- MCP (Model Context Protocol)
MCP (Model Context Protocol) is Anthropic’s protocol for exposing tools and resources to Claude. It is the most actively used of the…
- Mention Rate
Mention rate is the binary presence metric at L1 of the SOV pyramid: the proportion of AI answers that contain at least…
- Multimodal AI
Multimodal AI is artificial intelligence that processes and combines several input types — text, images, audio and sometimes video — in one…
N
- N-Sampling (N=5)
N-sampling is the per-query, per-engine sample size in a SOV run, default N=5: each query is asked five times on each engine…
P
- Parametric Memory
Parametric memory is knowledge encoded in a model's weights — answers given without any web search. SkuLift separates parametric SOV from web-grounded…
- Pillar Page
A pillar page is a comprehensive, authoritative page that covers a broad topic end-to-end and links out to a cluster of focused…
- PWC (Position-Weighted Count)
Position-Weighted Count (PWC) is the central L3 quality metric. PWC = Σ over brand sentences ( wcₛ · exp(−posₛ / |S|) )…
Q
- Query Intent
Query intent is the real goal behind a query — what the person actually wants — beyond the literal words. Engines infer…
R
- RAG (Retrieval-Augmented Generation)
Retrieval-Augmented Generation (RAG) is a pattern where an AI engine retrieves relevant documents first, then generates its answer grounded in those sources.…
S
- Schema.org
Schema.org is a shared vocabulary of types and properties — Organization, Product, FAQPage, DefinedTerm and more — that search engines and AI…
- Semantic Structuring
Semantic structuring is organising content so its meaning is explicit to machines — descriptive headings, answer-first sections, FAQs and structured data —…
- Sentiment Analysis
Sentiment analysis is the polarity of a brand mention — positive, neutral or negative — measured as an L3 quality component of…
- Share of Voice
Share of Voice is SkuLift’s core measurement metric — the full name of SOV. It is multi-engine across ChatGPT, Claude, Gemini and…
- SOV (Share of Voice)
Share of Voice (SOV) is SkuLift’s core measurement metric. It is multi-engine — ChatGPT, Claude, Gemini, Perplexity — and multi-level, structured as…
- SOV Pyramid
The SOV pyramid is a four-level decomposition of Share of Voice: L1 Presence (binary mention rate), L2 Volume (mention count + WCS),…
- Structured Data
Structured data is machine-readable markup that describes the meaning of a page’s content — its entities, types and relationships — so AI…
- Studio
A Studio is a table-shaped data workspace with typed columns, a lift status per row, and integration into the agentic loop. Studios…
T
- Type A (Branded Query)
A Type A query is a branded query: it contains the brand name. The person is already asking about you, so the…
- Type B (Unbranded Query)
A Type B query is an unbranded query: no brand name, just a category or generic need. The engine chooses who to…
- Type C (Comparative Query)
A Type C query is a comparative query — X versus Y, or best tool for a job. The engine weighs options…
W
- WCS (Word Count Share)
Word Count Share (WCS) is the L2 volume metric. It is computed at the sentence level: the words of every sentence containing…
- Web Grounding
Web grounding is when an AI engine bases its answer on a live web search rather than on its parametric memory alone…
- Wikidata
Wikidata is a free, collaborative, structured knowledge base of entities and their relationships. Search and AI engines read it to recognise and…
Z
- Zero-Click Search
Zero-click search is a search that resolves on the results page itself — through an AI Overview, snippet or panel — so…