GLOSSARY
The vocabulary of AI search, defined.
Every term you need to talk about Generative Engine Optimization with precision: what it means, and why it matters when AI answers questions about your brand.
- GEO (Generative Engine Optimization)
- The practice of making a website easier for generative AI engines to read, retrieve, and cite as they build their answers. GEO brings together technical crawlability, answer-ready content, explicit entities, structured data, and continuous measurement of whether models really mention and link the brand. Unlike classic SEO, the goal isn't a ranked position but inclusion as a trusted source inside a synthesized response.
- AEO (Answer Engine Optimization)
- Optimization aimed at answer engines, the systems that return a single synthesized answer instead of a list of links. AEO structures content as clear questions and direct answers, so an engine can lift a correct, attributable response. It overlaps heavily with GEO: AEO leans on the answer format, GEO on the full generative-retrieval surface.
- LLMO (Large Language Model Optimization)
- A near-synonym for GEO that frames the work as influencing what large language models say about an entity. LLMO covers both grounded answers (with live web retrieval) and a model's parametric knowledge, and it includes monitoring brand mentions across models, correcting misinformation, and reinforcing accurate, citable sources.
- AIO (AI Optimization)
- An umbrella term for optimizing a brand's presence across AI surfaces: assistants, overviews, and answer panels. AIO is often used interchangeably with GEO and AEO; in practice it points to a broad strategy spanning content, structured data, crawler access, and citation tracking, rather than a single tactic.
- Schema markup
- Structured data added to a page with the Schema.org vocabulary, so machines can understand what it means: that a string is a product, a price, an FAQ, or an organization. Search and AI systems use it to interpret content reliably and to power rich results and citations. It's typically expressed as JSON-LD.
- JSON-LD
- JSON for Linking Data: the recommended format for embedding Schema.org structured data in a page, placed in a <script type="application/ld+json"> block. It keeps machine-readable metadata separate from the visible HTML, which makes it easy to maintain and validate. Google and AI engines parse JSON-LD to understand entities, FAQs, and relationships.
- FAQPage
- A Schema.org type that marks up a list of questions and their answers. Adding FAQPage JSON-LD helps search and AI systems extract direct Q&A pairs from a page. Policies require the marked-up questions and answers to be visibly present on the page: invisible or mismatched FAQ markup violates structured-data guidelines.
- SpeakableSpecification
- A Schema.org property that flags the sections of a page best suited to be read aloud by voice assistants. By pointing out concise, self-contained passages, SpeakableSpecification helps voice and answer engines pick a clean snippet to speak, which also tends to line up with the passages AI engines quote.
- llms.txt
- A proposed plain-text file at a site's root that gives language models a curated map of the site: who you are, and which pages to read first. It works like a human-readable index built for AI, pointing models to the most authoritative, answer-ready content instead of leaving them to crawl blindly.
- GPTBot
- OpenAI's web crawler, used to gather content that may train and inform its models. Site owners control its access through robots.txt directives. Allowing GPTBot makes a site eligible to be read and potentially cited; blocking it removes the site from that pipeline.
- ClaudeBot
- Anthropic's web crawler for Claude. Like the other AI crawlers, its access is governed by robots.txt. Whether ClaudeBot can fetch a site decides if that content can inform Claude's grounded answers and citations.
- PerplexityBot
- The crawler Perplexity uses to index pages for its answer engine. Perplexity is citation-heavy by design and surfaces sources alongside answers, so allowing PerplexityBot is often the most direct path to being shown as a linked source.
- Google-Extended
- A robots.txt token that decides whether Google may use a site's content for its generative AI products and model training, separately from normal Search indexing. Disallowing Google-Extended can keep content out of AI features without affecting traditional ranking.
- Citation
- A source an AI engine references, and usually links, while producing an answer. Citations are the currency of GEO: being cited transfers the engine's trust to the brand and can drive a click, while being left out means a competitor's source is shown in your place.
- Grounding
- Connecting a model's answer to external, verifiable sources, typically via live web retrieval, instead of relying only on its trained memory. Grounded answers cite real pages: that's exactly the surface GEO targets, being among the sources a grounded response is built from.
- RAG (Retrieval-Augmented Generation)
- An architecture where a model retrieves relevant documents and uses them as context to generate an answer. RAG underpins most grounded AI search: content that's easy to retrieve, clearly chunked, and unambiguous is more likely to be picked as supporting context and cited.
- Web search tool
- The capability that lets an assistant query the live web during a conversation and fold fresh results into its answer. When the web search tool is active, answers reflect current pages and citations: it's the condition under which citation monitoring measures real visibility.
- AI Overview
- Google's AI-generated summary, shown above the traditional results for many queries, synthesizing information and linking sources. Appearing within an AI Overview, or among its cited links, is a high-visibility GEO outcome, because it sits at the top of the page, ahead of the classic blue links.