The AEO Dictionary: 43 Terms Every Answer Engine Optimizer Should Know

Answer Engine Optimization inherited half its vocabulary from traditional SEO and invented the other half from scratch, which means most people working in it today are fluent in neither. Some terms (schema markup, E-E-A-T) carry over unchanged. Others (citation rate, chunking, RAG) describe mechanics that simply didn't exist in the ten-blue-links era. And a few (AEO vs. GEO vs. LLMO) are near-synonyms different companies settled on independently, which is exactly why they keep getting used interchangeably in ways that confuse newcomers.

This glossary collects the terms that actually come up in day-to-day AEO and GEO work — not an exhaustive academic list, but the vocabulary you'll hit reading a tool's documentation, a vendor's pitch deck, or a technical audit. Terms are alphabetised so you can jump straight to the one you need; each definition is written to stand alone, so you shouldn't need to read five others first to understand it.

AEO (Answer Engine Optimization)

The practice of structuring content so AI-powered answer engines — ChatGPT, Perplexity, Gemini, Claude, Google's AI Overviews — can extract, trust and cite it directly inside a generated response, rather than only ranking it in a list of links a person has to click through.

AI Citation

An instance where an answer engine references or attributes a claim to a specific source or domain inside a generated answer. This is the core unit of measurement in AEO and GEO work — the equivalent of a keyword ranking in traditional SEO, except it happens inside the answer itself rather than on a results page.

AI Overviews

Google's AI-generated summary that appears above traditional search results for many queries, synthesising information from multiple sources into a single answer with citations attached. Functionally, it's Google's own answer engine sitting on top of its existing search index.

Answer Box

A structured, direct-answer UI element — in a search results page or an AI chat interface — that displays a concise response to a query without requiring a click-through. Often pulled from a single, well-structured source that answered the question in a form the system could lift cleanly.

Answer Engine

Any AI system that responds to a user's query with a synthesised, direct answer rather than a list of links: ChatGPT, Perplexity, Gemini, Claude, and Google's AI Overviews are the four most-tracked answer engines as of 2026.

BYOK (Bring Your Own Key)

A pricing model where a SaaS tool uses the customer's own API key for underlying AI model calls, instead of reselling model usage at a markup. It typically results in a lower total cost for the buyer, since they're paying the model provider's real rate rather than a resold, marked-up rate.

Chunking

The process of splitting a document into smaller, semantically coherent passages so a retrieval system or language model can process, embed, and retrieve each one individually. A poorly-chunked page — one long wall of text with no clear section breaks — is harder for a retrieval system to pull a clean, isolated answer from.

Citation Rate

The percentage of tracked prompts, over a given period, in which an AI engine cites or mentions a specific brand or domain in its response. The primary KPI most GEO tracking tools, including rankahead, report as a visibility score.

Content Freshness

A ranking and citation signal reflecting how recently a piece of content was published or meaningfully updated, used by both search and AI engines as a rough proxy for accuracy. Stale content on a fast-moving topic is both less likely to rank and less likely to get cited, even if it was authoritative when it was written.

Crawler (AI Crawler)

An automated bot operated by an AI company — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot and others — that fetches web pages either to use as model training data or as a live retrieval source when answering a query. Site owners can allow or block specific AI crawlers via robots.txt.

Direct Answer

A response format where content states the answer to a likely question plainly and immediately — typically in the first sentence of a section — rather than making the reader infer it from surrounding narrative. The single highest-leverage structural change most pages can make for AEO.

Embeddings

Numerical vector representations of text that capture semantic meaning, allowing a machine to compare how similar two pieces of content are without matching exact words. Embeddings are the foundation of how modern retrieval systems find relevant passages for a given query.

Entity

A distinct, named “thing” — a person, brand, product, place, or concept — that search and AI systems recognise and track independently of the exact words used to describe it at any given moment. Strong entity association is part of why some brands get cited even for queries that never mention their name directly.

E-E-A-T

Google's framework for content quality: Experience, Expertise, Authoritativeness and Trustworthiness. Originally a search-ranking concept, it's now widely used — informally, since no AI lab has published an exact equivalent — as a proxy for which sources are considered safe enough for an AI system to cite.

FAQ Schema

A structured data format (written in JSON-LD) that explicitly marks up question-and-answer pairs on a page, making it far easier for an answer engine to identify, isolate and extract a discrete answer without having to parse it out of prose.

Featured Snippet

A search result excerpt Google displays above standard listings, directly answering the query in a highlighted box. Widely considered the direct precursor to today's AI Overviews and answer engines — pages already optimised for featured snippets tend to be well-positioned for AEO too.

Generative Search

Search interfaces that generate a synthesised, written answer to a query rather than only listing links to click through. The umbrella behaviour that AEO and GEO both exist to influence, whether the interface is a dedicated chat app or a search engine's AI-generated summary panel.

GEO (Generative Engine Optimization)

A closely related discipline to AEO, describing the broader practice of optimising a brand's presence across generative AI systems as a whole — not just a single answer to a single question, but overall citation frequency and prominence across an entire category of queries and competitors.

GPTBot

OpenAI's web crawler, used to gather training data and, more recently, to retrieve live content for ChatGPT's browsing and search features. Like other AI crawlers, it respects — and can be explicitly allowed or disallowed via — a site's robots.txt file.

Hallucination

When an AI system generates a confident-sounding but factually incorrect or entirely fabricated claim. A key risk answer engines actively try to mitigate, which is exactly why clear, checkable, well-sourced content tends to be favoured for citation over vague or unverifiable claims.

Indexability

Whether a page can technically be crawled, processed and stored by a search or AI system in the first place — a prerequisite for either ranking or being cited at all. A page can have perfect content and still be invisible if it's blocked by robots.txt, marked noindex, or unreachable behind a login wall.

JSON-LD

The most common syntax for embedding structured data (schema.org markup) inside a web page's code. Read by search engines and, increasingly, by AI systems, to understand a page's content unambiguously rather than having to infer it from visible text alone.

Knowledge Graph

A structured database of entities and the relationships between them, used by search engines to answer factual queries directly and to disambiguate similarly-named entities — the system, for example, that knows which "Amazon" you mean based on the rest of your query.

LLM (Large Language Model)

The type of AI model — GPT, Claude, Gemini and others — trained on large volumes of text that powers modern answer engines and chat interfaces. When people talk about “optimising for AI,” they typically mean optimising for how an LLM-powered system retrieves and cites content.

llms.txt

An emerging, community-proposed convention: a plain-text file at a site's root that gives AI agents a curated summary of a site's most important pages, in the same spirit as an XML sitemap but written for a language model to read rather than a search indexer.

Passage Ranking

The process by which a search or retrieval system scores individual sections of a page — rather than the page as a whole — for relevance to a specific query. It's why a single well-structured paragraph deep in a long page can get cited even if the page's overall topic is broader.

Prompt

The text input a user gives to an AI system — the closest equivalent AEO has to a "keyword" in traditional SEO, and the basic unit that GEO tracking tools monitor for brand visibility.

Query Fan-Out

The process by which a model internally breaks a single user prompt into several related sub-queries before formulating its final answer. It means a page can get cited for a fan-out query it never directly targeted — one reason AI-citation patterns can look less predictable than traditional keyword rankings.

RAG (Retrieval-Augmented Generation)

An architecture where a language model retrieves relevant passages from an external source — live web content, a document store, a database — before generating its response, rather than relying purely on knowledge baked in during training. RAG is the mechanism behind most answer engines that can cite current, specific sources.

Rank vs. Cite

The key distinction AEO and GEO practitioners draw constantly: ranking means appearing in a list of search results; citing means being directly referenced inside a generated answer. A page can do either without the other, and measuring only rankings will systematically miss half the picture in 2026.

Robots.txt (AI Directives)

The standard file used to instruct crawlers which parts of a site they may access, increasingly used to explicitly allow or block specific AI crawlers by name — GPTBot, ClaudeBot, Google-Extended and others each respect their own directive line.

Schema Markup

Structured vocabulary, defined by the schema.org standard, added to a page's code that explicitly labels its content for machines — this is an Article, this is a Question, this is a Price — removing the ambiguity a model would otherwise have to resolve on its own.

Semantic Search

Search that matches queries to content based on meaning and intent rather than exact keyword overlap, made possible by embeddings. It’s why a page about “cutting monthly software costs” can surface for a query about “reducing SaaS spend” without sharing a single exact phrase.

SERP (Search Engine Results Page)

The traditional page of ranked links returned for a search query — increasingly supplemented, and for a growing share of queries partially replaced, by an AI-generated answer positioned above it.

Source Attribution

The practice, by an answer engine, of naming or linking the origin of a claim inside its generated response. This is the mechanism that turns an internal citation into something visible, clickable, and trackable from the outside.

Structured Data

Any content formatted in a predictable, machine-readable way — tables, schema markup, consistent heading hierarchies — that reduces the amount of ambiguity a model has to resolve before it can safely reuse the content in an answer.

Token

The basic unit of text a language model processes, roughly three-quarters of a word in English. Relevant to AEO mainly because it determines cost and context-window limits during generation — which indirectly shapes how much of a long page a model can realistically consider at once.

Topical Authority

A site's perceived depth of expertise on a specific subject, built through the breadth and consistency of its content on that topic over time. Used by both search engines and AI systems as a trust signal that can outweigh any single page's individual quality.

Training Data

The corpus of text a language model was trained on. Content published before a model's training cutoff may shape its baseline “knowledge” of a topic — separate and distinct from what it retrieves live via RAG when answering a specific, current query.

Vector Database

A database optimised to store and search embeddings by similarity rather than exact match. The infrastructure layer that powers retrieval in most RAG-based answer engines, letting them find semantically relevant passages in milliseconds across huge document collections.

Visibility Score

A composite metric, used by GEO tracking tools like rankahead, representing how often and how prominently a brand appears across a set of tracked prompts and AI engines — typically expressed on a 0–100 scale and tracked daily to show whether visibility is improving or declining.

Zero-Click Search

A search or AI interaction where the user gets their answer directly from the results page or chat response, without ever clicking through to a source website. The defining economic challenge of the AEO era: you can be the source of an answer and receive zero traffic for it, which is exactly why citation tracking — not just click tracking — has become necessary.

Zero-Shot

A model's ability to answer a query correctly without having been given specific examples of how to answer that exact type of question beforehand — relevant to AEO because it means a well-structured page can influence an answer even on a topic the model has never been explicitly trained or fine-tuned to handle.

Vocabulary shifts fast in a field this new — some of these terms will consolidate, some will fall out of use, and new ones will show up within the year. If you want to see several of them in action rather than just defined, rankahead's own dashboard tracks citation rate, visibility score and rank-vs-cite gaps daily across ChatGPT, Claude, Gemini and Perplexity — worth a look once the definitions above stop being abstract and start being something you want measured for your own brand.