GEO Glossary
20 essential terms for anyone measuring or improving brand visibility in ChatGPT, Claude, Gemini, and Perplexity. Alphabetically sorted, with related terms linked throughout.
Automated agents (GPTBot, ClaudeBot, PerplexityBot) that index web content for AI training and retrieval.
Major AI providers deploy their own crawlers to index the public web. GPTBot (OpenAI), ClaudeBot (Anthropic), and PerplexityBot all follow robots.txt rules but serve different purposes — some for training data collection, others for real-time retrieval augmentation. Allowing these bots access to your content is a foundational GEO infrastructure step.
How frequently and favorably a brand appears in AI-generated responses.
AI Visibility encompasses both quantitative presence (citation frequency) and qualitative positioning (how a brand is described, at what prominence, and in what context). A brand can have high citation frequency but low AI Visibility if it's consistently mentioned last, described inaccurately, or framed negatively. True AI Visibility requires both dimensions.
How often an AI engine mentions a specific brand when answering queries about its category.
Citation frequency is the most basic GEO metric — the raw count of how many times a brand appears across a standardized set of benchmark queries. It's measured per engine (ChatGPT, Claude, Gemini, Perplexity) and aggregated into an overall score. A brand with 40% citation frequency across 47 queries is mentioned in roughly 19 of those queries.
The web of authoritative sources that reference a brand, building its credibility for AI engines.
AI models learn about brands not just from a brand's own website, but from the ecosystem of sources that discuss, review, and cite that brand. A strong citation network includes Wikipedia/Wikidata entries, industry publications, comparison sites, press coverage, and authoritative directories like Crunchbase or LinkedIn. The consistency of brand descriptions across these sources is as important as the volume of citations.
How well-documented and trustworthy a brand appears across authoritative sources (Wikipedia, Wikidata, industry publications).
Entity Authority is a composite signal that reflects how confidently AI engines can represent a brand. High entity authority means the brand is described consistently across multiple independent, reliable sources — Wikipedia, Wikidata, industry publications, press coverage. Low entity authority means the AI must guess or interpolate, leading to inaccurate or missing representations. Entity authority is built over months through deliberate citation network development.
The discipline of systematically measuring and improving a brand's visibility in AI engines like ChatGPT, Claude, Gemini, and Perplexity.
GEO emerged as a distinct discipline in 2023-2024 as generative AI engines began capturing a meaningful share of discovery queries. Unlike SEO, which optimizes for ranking algorithms and click-through rates, GEO optimizes for how AI models represent brands in generated responses. The core GEO framework consists of three phases: measurement (establishing a baseline visibility score), optimization (structured data, entity building, citation network development), and monitoring (tracking score changes over time).
An AI system (like ChatGPT) that generates direct answers rather than returning a list of links.
Generative engines represent a fundamental shift from traditional search engines. Instead of returning ranked lists of links, they synthesize information into direct answers. This changes the discovery dynamic: users don't need to click through to evaluate options — the AI does the evaluation for them. For brands, this means the decision about whether to include or exclude them from a recommendation happens inside the AI model, not in the user's mind.
A structured database of interconnected entities used by Google and AI engines to understand relationships between brands, people, and concepts.
Knowledge graphs store facts as machine-readable triples (subject → predicate → object), enabling AI systems to reason about relationships between entities. Google's Knowledge Graph powers Knowledge Panels in search results and heavily influences Gemini's responses. Wikidata is the largest open knowledge graph, used across many AI training pipelines. For GEO, establishing a correct, well-linked entity in relevant knowledge graphs is a high-priority infrastructure task.
The underlying AI model powering engines like GPT-4, Claude 3.5, and Gemini 1.5.
Large Language Models are neural networks trained on vast text corpora to predict and generate text. Each LLM has different training data composition, training cutoff dates, and alignment fine-tuning — which means each model represents your brand differently. GPT-4, Claude 3.5 Sonnet, and Gemini 1.5 Pro can each have meaningfully different impressions of the same brand, making cross-engine consistency a key GEO challenge.
A plain text file on a website that explains the site's content to AI bots in a structured way.
Proposed as an emerging standard by Answer.AI, llms.txt is placed at the root of a website (e.g., yoursite.com/llms.txt) and provides AI crawlers with a concise, structured summary of the site's content and how the brand wants to be represented. Unlike robots.txt (which controls access) or Schema markup (which annotates individual pages), llms.txt provides a top-level context document. Real-time engines like Perplexity can use it as an authoritative self-description source.
Whether a brand is mentioned as "the best option", "an option", or "what some users prefer" — quality matters, not just frequency.
Not all mentions are equal. A brand cited as "the leading solution" receives significantly higher GEO value than the same brand cited as "one of many options some users consider." Positional citation tracks the quality dimension of AI visibility: first mention vs. last mention, recommended vs. acknowledged, positively framed vs. neutrally listed. In Visivily's scoring methodology, positional citation quality accounts for 35% of the total Visibility Score.
A system used by Perplexity and others that combines live web search with LLM generation.
RAG systems augment an LLM's base knowledge with real-time retrieved content. Perplexity is the most prominent example: it searches the web for each query, retrieves relevant pages, and uses that content to ground the LLM's response. For GEO, this means freshness matters — brands with current, indexed, high-authority pages get cited in RAG-based engines even if the underlying LLM's training data is stale. SEO fundamentals (page authority, recency, crawlability) directly support RAG-based GEO.
Machine-readable HTML annotations (JSON-LD format) that help AI engines and search engines understand page content.
Schema markup (implemented as JSON-LD in a <script> tag) provides structured, unambiguous metadata about a page's content. For GEO, Organization schema (with sameAs links to Wikidata, LinkedIn, Wikipedia), Article schema on blog posts, FAQPage schema for common questions, and Service schema for offerings are highest priority. Schema markup is the fastest GEO lever available — it can be deployed on your own site without external dependencies and takes effect as soon as AI crawlers re-index your pages.
How precisely and consistently a brand is defined for AI systems, without ambiguity.
Semantic clarity means AI engines can confidently place a brand in the correct category, with the correct differentiators, serving the correct audience. Lack of semantic clarity occurs when a brand's descriptions vary across sources — different categories, different value propositions, different target markets. AI models exposed to this inconsistency produce inconsistent outputs. Building semantic clarity requires a deliberate brand definition exercise followed by systematic propagation across all owned and third-party sources.
In GEO audits, results reported at p<0.05 to ensure observed visibility scores aren't due to random variation.
AI engines produce non-deterministic outputs — the same query can yield different responses across sessions due to temperature settings and model sampling. A single measurement is therefore unreliable. Rigorous GEO methodology requires multiple sessions per query and statistical testing to ensure observed visibility scores reflect genuine brand representation rather than random variation. Visivily runs each query across 3 independent sessions and reports results at p<0.05 significance.
A network of interconnected content pieces that establish deep expertise on a specific topic for AI engines.
A topical authority cluster is a group of closely related content pieces — pillar pages, supporting articles, FAQ sections, case studies — that collectively signal deep expertise in a domain. For GEO, topical authority clusters work because AI models learn expertise associations from training data. A brand with 15 highly interlinked, detailed articles on SaaS security is more likely to be cited as an authority in that domain than a brand with one broad article. This strategy works for both SEO and GEO.
The date beyond which an LLM has no training data — brands that emerged after this date may be unknown to that model.
Every LLM has a knowledge cutoff date after which no new training data was incorporated. GPT-4's cutoff is April 2023; Claude's varies by version. Brands that launched or significantly expanded after a model's cutoff date will be underrepresented or unknown to that model. This is why Perplexity (RAG-based, real-time) and ChatGPT (primarily training-data-based) can give dramatically different results for the same brand. GEO strategy must account for each engine's cutoff and recency characteristics.
A 0-100 composite metric measuring a brand's presence across AI engines, weighted by citation frequency, quality, and consistency.
The Visibility Score is Visivily's primary output metric. It combines three components: citation frequency (40%) — how often the brand is mentioned across benchmark queries; positional citation quality (35%) — how prominently and favorably the brand is mentioned; and contextual accuracy (25%) — whether the AI's description of the brand is correct and complete. Scores are calculated per engine and aggregated. A score above 70 indicates strong AI visibility; below 30 indicates significant underrepresentation.
An open, structured knowledge base heavily used in LLM training — a key GEO signal for entity authority.
Wikidata is a free, collaborative knowledge base operated by the Wikimedia Foundation, storing facts as structured triples. It serves as the backbone of Wikipedia's infoboxes and is heavily cited in LLM training data due to its structured, reliable format. A Wikidata entry for your organization, properly populated and linked to your Wikipedia page and official website, sends one of the strongest entity signals available to AI systems. Google's Knowledge Graph, which powers Gemini's entity recognition, draws substantially from Wikidata.
When a user gets their answer directly from an AI response without visiting any website — GEO determines who wins in this environment.
Zero-click represents the ultimate shift in discovery dynamics. In the traditional web, even rich snippets still drove some clicks. In AI-mediated discovery, a user asking ChatGPT for a software recommendation and receiving a confident, detailed answer may never visit any vendor website. The decision happens inside the conversation. For brands, this means the consequence of poor AI visibility is not just lower ranking — it's complete exclusion from the buyer's consideration set. GEO is fundamentally a response to the zero-click reality.
47 queries. 4 engines. A concrete visibility number and a 30-page action plan — delivered in 14 days.