GLOSSARY
GEO & AI Search Terminology
Plain-language definitions for the vocabulary of AI search, with the sources behind them.
100 pages
GEO (Generative Engine Optimization)
Content optimization targeting generative AI assistants like ChatGPT / Perplexity / AI Overview, aiming to have your brand cited and recommended by AI.
Updated 2026-08-06SEO vs GEO
SEO targets blue-link rankings and clicks; GEO targets being directly cited and recommended by AI.
Updated 2026-08-10RAG (Retrieval-Augmented Generation)
AI retrieves relevant content from external knowledge sources before generating answers — entering this retrieval pool is the core battleground of GEO.
Updated 2026-08-06AI Overview (AIO)
Google's AI-generated summary at the top of search results, aggregating content from multiple pages — being cited in AIO equals premium visibility.
Updated 2026-08-06Citeability
How friendly a piece of content is to AI citation in terms of structure, clarity and authority — higher Citeability means greater chance of being retrieved by LLMs.
Updated 2026-08-10Information Gain
Exclusive content on a page that AI cannot derive from other public sources — the core of content competitiveness in the GEO era.
Updated 2026-08-06AI Mention Rate (Visibility)
Of N target queries, the percentage where AI answers mention your brand — the core GEO KPI.
Updated 2026-08-10Cited Pages
How many distinct pages of yours AI has cited — reflecting the breadth of your AI-recognized content assets.
Updated 2026-08-06Authority Score
SEMRush's authority score (0-100) per domain, combining backlink quality, traffic, vertical relevance and other signals.
Updated 2026-08-06Average Position
Your site's average ranking position across all keywords on Google SERP — closer to 1 is better, within 10 is page 1.
Updated 2026-08-06G-Power (5-Dimension Score)
Our proprietary GEO scoring framework: V (Visibility) + D (Depth) + R (Recommendation) + C (Competitiveness) + A (Authority), weighted into one score.
Updated 2026-08-10Meta-Semantic Query
The underlying semantic intent beneath surface-level phrasing — one meta-semantic intent corresponds to 50-100 variant queries.
Updated 2026-08-10Buyer Journey (4 Stages)
The 4-stage purchase decision journey: Awareness → Interest → Decision → Purchase, with completely different questions at each stage.
Updated 2026-08-06TL;DR Module
'Too Long; Didn't Read' — a 3-4 bullet-point essence summary at the article top, highly friendly to AI crawling.
Updated 2026-08-06Schema Structured Data
Using schema.org markup to make page content machine-readable — both Google rich results and AI crawling rely on it.
Updated 2026-08-06VDRCA (Five-Dimension Diagnosis)
VDRCA breaks 'why was this recommended' into five separately measurable dimensions: visibility, mention depth, recommendation tendency, competitiveness and citeability.
Updated 2026-08-10Generative Engine
A generative engine synthesises retrieved content into an answer rather than returning a list of links.
Updated 2026-08-06Answer Engine
An answer engine returns an answer instead of a result list; optimising for one is called AEO.
Updated 2026-08-06Query Fan-Out
Query fan-out is when a system decomposes one user question into several sub-queries, retrieves for each, and then composes an answer.
Updated 2026-08-06Passage Retrieval
Passage retrieval matches and takes content at paragraph level rather than whole-page level.
Updated 2026-08-06Grounding
Grounding is when a model bases its answer on retrieved external content rather than parametric memory alone.
Updated 2026-08-06Hallucination
A hallucination is output that reads as plausible but has no factual basis.
Updated 2026-08-06Embedding
An embedding turns text into a numeric vector so that semantic closeness becomes a computable distance.
Updated 2026-08-06Vector Search
Vector search finds the most relevant content by semantic distance rather than keyword presence.
Updated 2026-08-06Zero-Click Search
A zero-click search is one where the user reads the results and clicks nothing.
Updated 2026-08-06AI Referral Traffic
AI referral traffic is visits that arrive by clicking through from an AI platform's answer.
Updated 2026-08-06Branded Search
A branded search is a query for the brand or product name itself.
Updated 2026-08-06Citation
A citation is the link or attribution in an AI answer pointing at a specific source.
Updated 2026-08-10Share of Answer
Share of answer is the proportion of citation sources across a question set's answers that belong to one domain.
Updated 2026-08-06Entity
An entity is a uniquely identifiable object in a knowledge system — a company, a product, a person.
Updated 2026-08-06Brand Entity
A brand entity is an AI system's understanding of a brand's identity: who it is, what it does, and how it differs.
Updated 2026-08-06Knowledge Graph
A knowledge graph stores entities and the relationships between them in structured form.
Updated 2026-08-06Information Consistency
Information consistency means a brand's name, description, location and capability statements match everywhere it appears.
Updated 2026-08-06Original Data
Original data is measurement only you can produce and that cannot be derived from other public sources.
Updated 2026-08-06Quotable Line
A quotable line is a single sentence that carries complete meaning and still holds when lifted out of context.
Updated 2026-08-06Semantic Density
Semantic density is how much useful information a given length of text carries.
Updated 2026-08-06Retrieval Crawler
A retrieval crawler fetches pages to answer a user's current question, and determines whether content can be cited at all.
Updated 2026-08-06Training Crawler
A training crawler fetches content for model training, which has no direct bearing on whether you can be cited.
Updated 2026-08-06AI Crawler Access
AI crawler access means allowing AI platform crawl requests through both robots.txt and the network layer.
Updated 2026-08-06robots.txt
robots.txt is a text file at the site root declaring which crawlers may fetch which paths.
Updated 2026-08-06llms.txt
llms.txt is a proposed root-level file listing site content for large language models.
Updated 2026-08-10Server-Side Rendering
Server-side rendering means the page content is produced on the server and returned as HTML, rather than assembled by scripts in the browser.
Updated 2026-08-06Prerendering
Prerendering generates a page as static HTML ahead of time so crawlers can take it directly.
Updated 2026-08-06Crawl Log
A crawl log is the server-side record of every fetch request, with user agent, path, status code and response size.
Updated 2026-08-06Structured Data
Structured data annotates a page's meaning in machine-readable form, most commonly schema.org JSON-LD.
Updated 2026-08-06JSON-LD
JSON-LD is the JSON-based syntax for embedding structured data in a page, and the recommended schema.org format.
Updated 2026-08-06DefinedTerm
DefinedTerm is the schema.org type for marking up a term definition, usually paired with DefinedTermSet.
Updated 2026-08-06FAQPage
FAQPage is the schema.org type for marking up question-and-answer content.
Updated 2026-08-06Sitemap
A sitemap is an XML file listing indexable URLs so search engines and crawlers can discover content.
Updated 2026-08-06Canonical
A canonical declares the preferred URL for one piece of content, consolidating signals from duplicate addresses.
Updated 2026-08-06Duplicate Content
Duplicate content is one piece of content living at several URLs with no consolidation to a preferred address.
Updated 2026-08-06Slug
A slug is the URL segment identifying a specific page, usually derived from the title.
Updated 2026-08-06hreflang
hreflang declares the correspondence between language versions of the same content.
Updated 2026-08-06Meta-Semantic Question Pool
A meta-semantic question pool is the set of a few hundred heat-weighted decision questions that thousands of surface phrasings collapse into.
Updated 2026-08-06Prompt Set
A prompt set is the fixed group of questions used to measure AI visibility, and the denominator behind every related metric.
Updated 2026-08-06Long-Tail Keyword
A long-tail keyword is a specific query with low volume but clear intent.
Updated 2026-08-06Head Term
A head term is a high-volume, highly competitive query with relatively broad intent.
Updated 2026-08-06Content Cluster
A content cluster is a hub page plus interlinked detail pages built around one core topic.
Updated 2026-08-06Internal Linking
An internal linking matrix is a designed set of links between pages, as opposed to links added ad hoc.
Updated 2026-08-06Buyer Journey Stage
A buyer journey stage is where someone sits between first awareness and purchase, which determines what they ask.
Updated 2026-08-06AEO (Answer Engine Optimization)
AEO is optimisation for retrieval systems that answer directly, overlapping heavily with GEO.
Updated 2026-08-10LLMO (LLM Optimization)
LLMO frames the work as optimising for large language models, a phrasing used mainly in engineering circles.
Updated 2026-08-06Google AI Mode
AI Mode is the conversational, synthesised-answer mode inside Google Search.
Updated 2026-08-06Mention Depth
Mention depth measures how substantively a brand is discussed in an answer — merely listed, or given reasons and use cases.
Updated 2026-08-06Recommendation Tendency
Recommendation tendency measures whether the AI is recommending the brand or merely listing it as an example.
Updated 2026-08-06Index
Indexing is when a page enters the set a search or retrieval system can draw from.
Updated 2026-08-06Crawl Budget
Crawl budget is how many requests a crawler is willing to spend on a site in a given period.
Updated 2026-08-06noindex
noindex tells crawlers not to index a page.
Updated 2026-08-06301 Redirect
A 301 signals that a resource has moved permanently and passes the old address's signals to the new one.
Updated 2026-08-06410 Gone
A 410 states the resource has been permanently removed — a stronger signal than 404.
Updated 2026-08-06Thin Content
Thin content is a page with little information, high similarity to others, or that merely rephrases something else.
Updated 2026-08-06Core Web Vitals
Core Web Vitals are Google's page experience metrics: LCP, INP and CLS.
Updated 2026-08-06Time to First Byte
TTFB is the time from issuing a request to receiving the first byte.
Updated 2026-08-06CDN and Edge Layer
A CDN or edge layer handles requests before the origin and can block outside your site configuration.
Updated 2026-08-06User Agent
A user agent is the identity a client declares in its HTTP headers.
Updated 2026-08-06SERP
A SERP is the results page a search engine returns.
Updated 2026-08-06Featured Snippet
A featured snippet is the block at the top of results that answers directly.
Updated 2026-08-06People Also Ask
People Also Ask is the expandable set of related questions in search results.
Updated 2026-08-06Brand SERP
A brand SERP is the results page returned when someone searches the brand name.
Updated 2026-08-06Knowledge Panel
A knowledge panel is the sidebar block presenting an entity's core information.
Updated 2026-08-06E-E-A-T
E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness.
Updated 2026-08-06Topical Authority
Topical authority is how credible a site is considered on a given subject.
Updated 2026-08-06Referring Domain
Referring domains are the count of distinct domains linking to a site.
Updated 2026-08-06Digital PR
Digital PR is earning third-party mentions and links through content worth reporting.
Updated 2026-08-06User-Generated Content
UGC is content produced by users rather than the brand — community discussion, reviews and similar.
Updated 2026-08-06B2B Review Site
A B2B review site is a third-party platform collecting business software reviews.
Updated 2026-08-06Cited Source Mix
Cited source mix is the distribution of source types cited across a batch of answers.
Updated 2026-08-06Wikidata
Wikidata is an open structured knowledge base of entities and their properties.
Updated 2026-08-06Large Language Model
An LLM is a model trained on large text corpora that generates coherent text.
Updated 2026-08-06Context Window
A context window is the maximum amount of text a model can process at once.
Updated 2026-08-06Chunking
Chunking splits a long document into passages so each can be embedded and retrieved separately.
Updated 2026-08-06Reranking
Reranking applies a second relevance ordering to an initially retrieved set.
Updated 2026-08-06Freshness
Freshness is a content's recency and whether the system treats it as currently valid.
Updated 2026-08-06Multi-Source Agreement
Multi-source agreement is when the same claim is corroborated across several independent sources.
Updated 2026-08-06Impression
An impression is a count of the content appearing in a result set.
Updated 2026-08-06Click-Through Rate
Click-through rate is clicks divided by impressions.
Updated 2026-08-06Position Band
A position band groups rankings into ranges — 1–3, 4–10, 11–20, 21+ — to read yield.
Updated 2026-08-06Baseline
A baseline is the starting measurement taken before optimisation, used for later comparison.
Updated 2026-08-06Attribution
Attribution is determining which channel produced a visit or conversion.
Updated 2026-08-06AI Visibility
AI visibility is made up of four separately measurable parts — appearance, description, recommendation and support. Reporting only appearance flattens three different situations into one percentage.
Updated 2026-08-10