Answer engine optimization borrows words from search, AI research and marketing measurement, and mixes them freely. This glossary defines the 64 terms you're most likely to meet, in plain English, grouped by theme. Bookmark it, share it with stakeholders, and use it to agree on definitions before you agree on targets.
Core concepts
Answer engine. Any system that responds to a question with a direct answer rather than a list of links. Includes ChatGPT, Perplexity, Gemini, Claude, Copilot, Grok, Google AI Overviews and AI Mode, and voice assistants.
Answer engine optimization (AEO). Work to get your brand mentioned, accurately described and cited inside answers from answer engines. See what answer engine optimization is.
Generative engine optimization (GEO). A near-synonym for AEO, coined in a 2023 research paper, focused on AI systems that generate answers from retrieved sources. See AEO vs GEO vs LLMO.
LLM optimization (LLMO). Improving how large language models represent and mention your brand; emphasizes the training data side.
AI search. Search experiences where AI generates the answer: ChatGPT search, Perplexity, Google AI Overviews and AI Mode, Copilot and others.
AI Overviews. AI-generated summaries at the top of some Google results, with links to sources.
AI Mode. Google's conversational search experience, where the whole result is an AI answer with follow-up questions.
Zero-click search. A search where the user's need is met without clicking through to a website. See zero-click answers.
Entity. A distinct, identifiable thing such as a company, product, person or place, as understood by search engines and AI models.
Entity consistency. Describing your brand the same way everywhere, so systems are confident all mentions refer to the same entity.
How AI models work
Large language model (LLM). A model trained on large amounts of text to predict and generate language. Powers ChatGPT, Claude, Gemini and others.
Training data. The text a model learned from. Shapes the model's background knowledge.
Training cutoff. The date after which a model has no knowledge from training data.
Parameters / weights. The internal values a model learns in training. Brands can't edit them directly.
Token. A chunk of text, roughly a word or part of a word, that models read and write.
Context window. How much text a model can consider at once, including retrieved documents.
Hallucination. A confident but false statement generated by a model. See AI visibility metrics for tracking accuracy.
Grounding. Basing an answer on retrieved sources rather than model memory alone.
Fine-tuning. Further training a model on specific data. Relevant for companies building their own AI, not usually for AEO.
System prompt. Hidden instructions a platform gives its model about how to behave. Not something brands can influence.
Retrieval and search
Retrieval-augmented generation (RAG). Looking up relevant documents and then generating an answer from them. See RAG explained for marketers.
Index. A searchable database of crawled pages. AI assistants use their own indexes, partners' (such as Bing or Google), or both.
Query rewriting. When an AI turns a user's question into one or more search queries.
Query fan-out. Splitting one question into many related searches and combining the results.
Chunk / passage. A section of a page that retrieval systems score and select.
Chunking. Splitting documents into passages for retrieval.
Semantic search. Finding text by meaning rather than exact words, often using embeddings.
Embedding. A numerical representation of text meaning, used to compare passages with queries.
Reranking. A second step that reorders retrieved candidates by relevance.
Freshness. How recent content is; weighted more heavily for time-sensitive questions.
Crawlers and access
Crawler / bot. Software that fetches web pages automatically.
Training crawler. Collects content that may be used to train models (for example GPTBot, ClaudeBot, CCBot).
Search crawler. Builds an index for AI search answers (for example OAI-SearchBot, Claude-SearchBot, PerplexityBot).
User-triggered fetcher. Loads a page because a user asked (for example ChatGPT-User, Claude-User, Perplexity-User).
Control token. A robots.txt user agent that controls AI use without being a separate crawler (for example Google-Extended, Applebot-Extended).
robots.txt. A file telling crawlers what they may fetch. See our AI crawlers and robots.txt guide.
Server-side rendering (SSR). Generating a page's HTML on the server, so crawlers that don't run JavaScript can read it.
Prerendering. Generating static HTML for pages ahead of time; an alternative to SSR.
Structured data / schema markup. Machine-readable descriptions of page content, usually JSON-LD. See schema markup for AEO.
llms.txt. A proposed file listing a site's key content for AI systems. Not confirmed as a ranking or citation factor by major providers.
IndexNow. A protocol for notifying participating search engines, including Bing, that pages changed.
Measurement
Prompt. A question or instruction given to an AI assistant.
Prompt set. A fixed list of buyer questions tracked over time.
Mention. An AI answer naming your brand.
Mention rate. Share of answers that mention you.
Share of voice (SoV). Your mentions as a share of all mentions of you and tracked competitors.
Position / prominence. Where in an answer you appear and whether you're the main recommendation.
Sentiment. Whether you're described positively, neutrally or negatively.
Framing. The specific qualifiers attached to your brand ("best for enterprises," "expensive").
Citation. A link to a source in an AI answer.
Citation share. Your domain's share of citations across your prompts.
Cited domains. The third-party sites AI relies on for your category.
AI referral traffic. Visits from links in AI answers (chatgpt.com, perplexity.ai and others).
Self-reported attribution. Asking customers how they heard about you, including "AI assistant."
Answer variability. The way AI answers differ between runs; why measurement uses repeated runs.
Strategy and tactics
Answer-first writing. Opening each section with a direct answer. See how to write content AI assistants cite.
Use-case page. A page explaining how a product serves a specific audience or job.
Comparison page. A page comparing your product with alternatives, usually with a table.
Roundup / listicle. An article listing the "best" options in a category; heavily cited for recommendation questions.
Digital PR. Earning coverage and mentions in third-party publications.
Review program. A steady process for collecting customer reviews on relevant platforms.
Agentic browsing. AI agents that browse websites and take actions on a user's behalf, such as comparing prices or filling forms.
AI shopping. Product discovery and recommendations inside AI assistants, often using product feeds and structured data.
For quick definitions, see our AEO & GEO glossary.
The bottom line
Shared vocabulary makes AEO easier to plan and report. Agree on what you mean by mention rate, share of voice and citation share, keep the distinction between training and retrieval clear, and know which crawler does what. Everything else builds on those definitions.
Want to go deeper than definitions? Follow our free AEO and GEO curriculum.
