GEO Glossary

The GEO glossary by Vizible AI: clear, citable definitions of the terms that decide your AI visibility, from GEO and AEO to citations and AI crawlers.

• AI Referral Traffic

  AI referral traffic is the share of a site's traffic arriving specifically from AI platforms, ChatGPT, Perplexity, Gemini, and similar tools, as opposed to traditional organic or paid search, typically identified in analytics by referrer domains tied to those platforms.

• AI Share of Voice (AI SOV)

  AI Share of Voice (AI SOV) is the percentage of AI-generated responses that mention a brand, out of all AI responses to a defined set of relevant industry queries, benchmarked against how often competitors get mentioned for the same queries.

• AI Visibility

  AI visibility is whether a brand appears at all in AI-generated answers for a defined set of relevant queries, regardless of whether that appearance rises to the level of a formal citation.

• AI Visibility Score

  An AI visibility score is a single composite number summarizing how discoverable and citable a brand is across AI systems, typically rolling up mention rate, citation rate, position, and sentiment into one comparable figure.

• AIO Trigger Rate

  AIO trigger rate is the fraction of tracked queries that cause Google to display an AI Overview at all, before any question of who gets cited within it even applies.

• Agentic Commerce

  Agentic commerce is AI agents completing purchases autonomously on a user's behalf, from product research through checkout, without a human directly clicking through each step, an emerging capability adjacent to GEO as brands increasingly need product data legible to purchasing agents, not just answer engines.

• Agentic Search

  Agentic search happens when an AI agent researches and selects information on a user's behalf: running its own sub-queries, evaluating multiple sources, and delivering a synthesized recommendation, rather than a human directly reading and judging a results page themselves.

• Amazonbot

  Amazonbot is Amazon's web crawler, feeding content into Alexa's answer capabilities and Amazon's other AI-driven products, giving it a narrower but growing relevance as voice and shopping-assistant answers increasingly draw on web content rather than only Amazon's own catalog and review data.

• Answer Block / Atomic Answer

  An answer block, or atomic answer, is a self-contained passage, typically 40 to 100 words, that fully answers a specific question on its own, without requiring a reader or an AI system to reference surrounding paragraphs to understand it.

• Answer Engine

  An answer engine is an AI-powered system, like ChatGPT, Perplexity, or Google AI Overviews, that responds to a query with a direct, synthesized answer and cited sources, rather than a ranked list of links a user has to click through to find their own answer.

• Answer Engine Optimization (AEO)

  Answer Engine Optimization (AEO) is the practice of structuring content so an AI assistant can read it, verify it, and name it directly in the answer it gives a user. It describes functionally the same goal as GEO, with roots more in commercial SEO tooling than in academic research.

• Answer Position

  Answer position is where a brand's mention or citation falls within an AI-generated response, whether it's named first or last, and where it ranks within the list of cited sources shown alongside the answer.

• Applebot-Extended

  Applebot-Extended is Apple's opt-out token, used in robots.txt, that lets a site exclude its content from training Apple Intelligence and Apple's other generative AI features, separate from the standard Applebot rules governing Apple's search and Siri-related indexing across the web.

• Article Schema

  Article schema is a structured-data type that captures an editorial page's author, publish date, headline, and related metadata in machine-readable form, giving crawlers explicit confirmation of authorship and freshness rather than requiring them to infer these details from visible page text.

• Attribution Gap

  The attribution gap is the numerical difference between how many relevant sources an AI system actually consults while forming an answer and how many of those sources it visibly cites in the response a user sees.

• Authoritative Voice

  Authoritative voice is the Princeton GEO paper's tested tactic of writing with confidence and eliminating hedging language, which the study measured as producing a smaller but still measurable 10 to 20 percent improvement in citation visibility, the mildest effect among the paper's positively performing tactics.

• BLUF (Bottom Line Up Front)

  BLUF, or Bottom Line Up Front, is a writing structure that states the direct answer to a question first, before any supporting context, background, or qualifiers, reversing the traditional build-up-to-the-point structure common in narrative and academic writing styles.

• Brand Mention

  A brand mention is any reference to a brand inside an AI-generated answer that isn't a formal, linked citation, the model naming a company without attributing a specific claim or a source URL to it.

• Bytespider

  Bytespider is ByteDance's web crawler, collecting training data across the web for its AI products, and it has a documented reputation for frequently ignoring robots.txt disallow rules that would normally block it, unlike most other major AI crawlers currently in operation.

• CCBot

  CCBot is the crawler behind Common Crawl, a large public dataset of web pages that isn't tied to any single AI lab but feeds model training for many different companies, making it one of the most consequential crawlers to account for even though it isn't operated by a household-name AI product.

• Chat Referral Traffic

  Chat referral traffic is visits arriving from links shared directly within an AI chat interface, such as a link a user clicks inside a ChatGPT or Claude conversation, forming a distinct analytics category from both traditional search referral and general AI referral traffic from a platform's search feature.

• ChatGPT Search

  ChatGPT Search is OpenAI's web-grounded answer mode inside ChatGPT, which retrieves current web pages before generating a response and displays a source panel alongside the answer. It launched as a standalone feature in late 2024 and uses a separate crawler, OAI-SearchBot, to build its citation index apart from GPTBot's model-training crawl.

• ChatGPT-User

  ChatGPT-User is the user-agent OpenAI applies when ChatGPT fetches a specific page live, in real time, at a user's explicit request within a conversation, distinct from the background indexing done by GPTBot or OAI-SearchBot. It represents an on-demand fetch rather than a systematic crawl.

• Chunking

  Chunking is how AI systems split content into smaller passages, typically 200 to 500 tokens, before indexing and retrieving it, meaning a page is rarely evaluated as one continuous whole but as a set of independent segments competing separately for retrieval.

• Citability

  Citability is how likely an AI engine is to actually quote or attribute a specific passage when constructing its answer, distinct from retrievability, since a passage can be retrieved into an engine's candidate set without ever being judged specific or self-contained enough to quote.

• Citation Accuracy

  Citation accuracy is whether an AI system's description of a brand, in the moments it does mention or cite that brand, is actually factually correct, rather than outdated, garbled, or confused with a competitor.

• Citation Gap

  A citation gap is a query where a brand ranks well in traditional search results but its content is omitted or misrepresented in the AI-generated answer for that same query.

• Citation Panel

  A citation panel is the list of sources an AI engine displays alongside a generated answer, typically as numbered links or a sidebar, separate from any inline citations embedded in the answer text itself. Appearing in a citation panel without an inline citation still counts as a mention, but carries less prominence than being cited directly within the answer.

• Citation Rate

  Citation rate is the percentage of relevant AI-generated answers that cite a brand's domain as a source, out of every answer where citing that brand would have been relevant. It's the most commonly cited single performance metric across the GEO and AEO industry.

• Citation Share

  Citation share is a brand's percentage of citations within an AI answer, normalized against the total number of citation slots that answer actually contains, rather than simply counting whether a citation happened at all.

• Citation-Worthiness Scoring

  Citation-worthiness scoring is a secondary filtering layer some AI engines apply to already-retrieved results, beyond standard search ranking, to decide which sources actually earn a citation in the final answer. A page can rank well and still get retrieved without ever being judged citation-worthy enough to appear in the answer itself.

• Cite Sources (GEO method)

  Cite sources is one of the optimization tactics the Princeton GEO paper tested directly: attributing specific claims to authoritative external sources within the content itself, which the study measured as producing roughly a 30 to 40 percent improvement in citation visibility over an unoptimized baseline.

• Claude

  Claude is Anthropic's model family, which added a web-grounded search mode that retrieves live pages and cites them alongside a generated answer, structurally similar to ChatGPT Search. Anthropic runs two separate crawlers for it: ClaudeBot collects training data, while Claude-SearchBot builds the separate index that Claude's search feature actually cites from.

• Claude-SearchBot

  Claude-SearchBot is Anthropic's crawler dedicated to indexing web content for Claude's search and citation feature, separate from ClaudeBot, which collects training data. It builds the pool of retrievable, citable pages Claude draws on when it grounds an answer in live web sources.

• ClaudeBot

  ClaudeBot is Anthropic's crawler for collecting web content used in training its Claude models, operating separately from Claude-SearchBot, the crawler that builds the index Claude's web search and citation feature actually draws from. As with OpenAI's crawlers, blocking one does not automatically block the other.

• Context Window

  A context window is the amount of text an AI model can process at once, in a single conversation or request, which directly limits how much of a given page, or how many pages, the model can actually see and draw from when generating a grounded, cited answer.

• Conversational Search

  Conversational search is a natural-language search interaction where a user asks a question the way they'd ask a person, and receives one synthesized answer instead of a page of ranked links. It's the interface pattern behind ChatGPT, Perplexity, and Google's AI Mode.

• Copilot (Bing Copilot)

  Copilot is Microsoft's AI-search surface embedded in Bing and Edge, blending conversational answers with traditional Bing search results and citations. It draws on Bing's existing index and ranking signals, giving it a retrieval profile closer to conventional search authority weighting than to Perplexity's freshness-driven approach.

• Crawl-to-Refer Ratio

  Crawl-to-refer ratio is a measure of how many pages an AI system consumes through crawling versus how many outbound referrals it actually generates back to the site, a ratio that varies sharply across platforms, some of which crawl heavily while sending very little traffic back.

• Dark AI Traffic

  Dark AI traffic is AI-driven site visits that arrive without referrer headers, making them invisible to standard analytics and typically misclassified as direct traffic instead of being correctly attributed to the AI platform that actually sent the visitor.

• E-E-A-T

  E-E-A-T is Google's content-quality framework, Experience, Expertise, Authoritativeness, Trustworthiness, originally built for human search-quality raters, that now functions informally as a trust filter influencing whether AI systems treat a source as credible enough to cite.

• Editorial Authority Weighting

  Editorial authority weighting is ChatGPT Search's tendency to concentrate citations among established media outlets and sites with long-standing editorial reputations, echoing the trust signals traditional search ranking has relied on for years. A page's accumulated domain authority, not just how directly it answers a query, materially shapes whether ChatGPT Search cites it.

• Embeddings

  Embeddings are numerical representations of text that let AI systems measure semantic similarity between a query and a set of candidate passages, forming the underlying mechanism that makes semantic search and most retrieval-augmented generation possible in the first place.

• Entity

  An entity is a distinct, recognizable thing, a brand, person, product, or concept, that an AI system can identify, disambiguate, and connect consistently across different sources, rather than treating as just an unrelated string of text each time it appears.

• Entity Chain

  An entity chain is the distributed set of independently verifiable mentions and structured data about a brand, spanning its own site, Wikipedia, LinkedIn, industry directories, and press coverage, that any retrieval system can resolve back to one consistent entity. The stronger and more consistent this chain, the more confidently an AI system can confirm what a brand is.

• Entity Salience

  Entity salience is a score, typically between 0 and 1, indicating how prominent or central a given entity is within a piece of text, a concept borrowed from natural-language-processing tools like Google's Cloud Natural Language API.

• Entity-Level Verification

  Entity-level verification is Gemini's practice of cross-checking a source's claims against Google's knowledge graph before citing it, confirming that the brand, person, or fact referenced is a recognized, disambiguated entity rather than an unverified mention. A claim tied to a well-established entity clears this check more easily than one tied to an ambiguous or newly formed one.

• Equalizer Effect

  The equalizer effect is the Princeton GEO paper's finding that citations let lower-ranked sources, such as a page sitting at position five in traditional search, achieve outsized visibility gains within an AI-generated answer relative to their organic rank, a genuinely counter-intuitive result given how strongly position predicts visibility in traditional search.

• Extractability

  Extractability is how cleanly a chunk of content can be lifted and reused by an AI system without needing surrounding context to remain accurate or coherent, functioning as a practical measure of whether a passage would still make sense if it were the only thing quoted.

• FAQ Schema

  FAQ schema is structured markup for question-and-answer pairs on a page, explicitly labeling each question and its corresponding answer so a crawler doesn't have to infer that structure from formatting alone. It's one of the most directly applicable schema types for content already organized around specific questions.

• Fluency Optimization

  Fluency optimization is the Princeton GEO paper's tested tactic of rewriting content for readability without changing its underlying claims, which the study measured as producing a moderate 15 to 30 percent improvement on its own, and a stronger effect when combined with statistics addition.

• Freshness Signal

  A freshness signal is a visible cue, such as a dated last-updated line, current statistics, or recent examples, that confirms a page's content is current rather than stale, giving both traditional search and AI retrieval systems a basis for judging recency independent of publish date.

• GEO-bench

  GEO-bench is the benchmark the Princeton and IIT Delhi GEO paper built to test and compare optimization tactics, consisting of 10,000 queries spanning 8 different domains, used as the standardized dataset every tactic in the study, from statistics addition to keyword stuffing, was measured against.

• GPTBot

  GPTBot is OpenAI's crawler for collecting web content used to train its models, operating separately from the crawlers OpenAI uses to power ChatGPT Search citations. A site can block GPTBot specifically to opt out of having its content used for model training, without affecting whether ChatGPT Search can still retrieve and cite that same content.

• Gemini

  Gemini is Google's family of AI models, powering both Google AI Overviews and a standalone chat product with its own citation behavior, plus a separate Deep Research agent mode that compiles longer, multi-source reports. Its retrieval is search-grounded: it connects a generated answer back to Google's own search index before applying a further citation filter on top.

• Generative Engine

  A generative engine is an AI system, such as ChatGPT, Perplexity, or Google's AI Overviews, that answers a query by synthesizing one direct response from multiple retrieved sources, instead of returning a ranked list of links for the user to click through themselves.

• Generative Engine Optimization (GEO)

  Generative Engine Optimization (GEO) is the practice of structuring and positioning content so generative AI systems, including ChatGPT, Perplexity, Gemini, and Claude, retrieve it, quote it, and cite it when answering a question, rather than optimizing for a ranked list of links the way traditional SEO does.

• Google AI Mode

  Google AI Mode is a separate, fully conversational search interface within Google, distinct from AI Overviews, that lets a user ask follow-up questions in a multi-turn dialogue instead of issuing a single query. It functions as Google's more direct answer to ChatGPT Search and Perplexity, running deeper multi-step retrieval than the AI Overviews shown on standard results pages.

• Google AI Overviews (AIO)

  Google AI Overviews are AI-generated summary blocks that appear above traditional organic results for a large share of Google searches, synthesizing an answer from multiple cited sources. Whether an AI Overview appears at all for a given query, its trigger rate, is a separate variable from whether any specific brand gets cited within it once it does appear.

• Google-Extended

  Google-Extended is not a crawler itself, but a robots.txt token site owners use to opt specific content out of use by Gemini and Google's other generative AI products, separate from the standard Googlebot rules governing traditional search indexing. A page can stay fully indexed in Google Search while excluded from Gemini's training data.

• Grounding

  Grounding is when an AI system anchors its generated response in specific retrieved source material, a live web page, a document, a database record, rather than relying only on patterns learned during training. It's what separates a cited answer from an unsupported one.

• Hallucination

  A hallucination is when an AI system generates a plausible-sounding but factually unsupported claim, a wrong statistic, a misattributed quote, a fabricated detail, that isn't actually backed by any of the sources it retrieved when forming its answer.

• Hub and Spoke

  Hub and spoke is a content structure built around one page covering a core concept in depth, the hub, linked to multiple pages each addressing a related sub-question, the spokes, giving both crawlers and readers a clear map of how a topic's pieces relate to the whole.

• Information Gain

  Information gain is the unique value a piece of content adds beyond what's already publicly available on a topic, whether through original data, a distinct analytical angle, or details not covered by existing sources, rather than a restatement of consensus information.

• Inline Citation

  An inline citation is a footnote-style link embedded directly within an AI-generated answer's text, attributing one specific claim to one specific source, the way Perplexity does for nearly every statement it makes. Inline citations carry more weight than a general citation-panel listing, since they tie a brand's authority to an exact claim, not the answer as a whole.

• JSON-LD

  JSON-LD is the specific structured-data format, a block of JSON embedded in a page's HTML, that both traditional search engines and AI systems prefer for schema markup over older formats like microdata, because it's easier to generate, validate, and parse without altering a page's visible layout.

• Keyword Stuffing (anti-pattern)

  Keyword stuffing, as tested by the Princeton GEO paper, is the practice of repeating target keywords or phrases unnaturally throughout content, and notably the one tactic the study measured as performing worse than an unoptimized baseline on Perplexity specifically, actively reducing citation visibility.

• Knowledge Graph

  A knowledge graph is a structured database of entities and the relationships between them, Google's and Wikidata's are the two most influential, that AI systems consult to verify claims and resolve which real-world thing a name actually refers to.

• Large Language Model (LLM)

  A large language model (LLM) is an AI model trained on vast amounts of text to understand and generate human language. It's the underlying technology behind ChatGPT (OpenAI's GPT models), Claude (Anthropic), and Gemini (Google), and what every generative engine is built on top of.

• Mention Rate

  Mention rate is how often a brand appears in AI-generated responses across a defined set of relevant queries, expressed as a percentage, independent of whether that appearance counts as a formal, linked citation.

• Multimodal Search

  Multimodal search is search that combines text with images, voice, or video within a single query, letting a user, for example, submit a photo alongside a text question rather than describing what they're looking for in words alone, expanding what counts as a searchable input.

• OAI-SearchBot

  OAI-SearchBot is OpenAI's separate crawler that indexes web content specifically for ChatGPT Search, building the retrieval pool that feature cites from, independent of the GPTBot crawler used for model-training data collection. Blocking OAI-SearchBot removes a page from ChatGPT Search's citation pool without affecting training opt-outs.

• Organization Schema

  Organization schema is a structured-data type defining a brand's core entity information, including its official name, logo, founding date, and related identifiers, in machine-readable form that anchors every other page on the site to one consistent, verifiable organizational identity.

• Passage Retrieval

  Passage retrieval is the step where an AI system selects specific relevant text chunks, not entire pages, to build its answer, meaning a page can be retrieved for one passage while the rest of it never enters the AI system's consideration for that query at all.

• Perplexity

  Perplexity is an AI-native search engine built around inline, numbered citations, where nearly every claim in its answers links directly to a specific source. It positions itself explicitly as a citation-first answer engine, and its retrieval favors recently published, directly responsive pages over pages with heavier accumulated editorial authority.

• PerplexityBot

  PerplexityBot is Perplexity's crawler, responsible for indexing pages into the source pool its answer engine retrieves from and cites, functioning as the crawler equivalent of Perplexity's real-time, citation-first retrieval approach. Unlike some competitors, Perplexity runs one primary crawler rather than separating training and citation functions.

• Position-Adjusted Word Count (PAWC)

  Position-Adjusted Word Count, or PAWC, is the metric the Princeton and IIT Delhi GEO paper (arXiv 2311.09735) uses to score a source's visibility within an AI-generated answer, weighting how many of that source's words appear in the response and where they fall, with earlier-appearing content weighted more heavily.

• Prompt Monitoring

  Prompt monitoring is the practice of tracking how AI systems respond to a defined, recurring set of prompts over time, running the same questions on a schedule to see how mentions, citations, and sentiment shift.

• Query Coverage

  Query coverage is the size and scope, topical and geographic, of the query set used to measure any AI-visibility metric, and it directly determines how representative that measurement actually is.

• Query Fan-Out

  Query fan-out is when an AI system decomposes one user query into several related sub-queries, retrieving source material for each before synthesizing them into a single answer, rather than treating the original question as one atomic search against one index.

• Quotation Addition

  Quotation addition is the Princeton GEO paper's tested tactic of inserting direct quotes from named experts into content, which the study measured as producing roughly a 28 to 40 percent improvement in citation visibility, making it the second-strongest tactic tested after statistics addition.

• Real-Time Retrieval

  Real-time retrieval is Perplexity's defining retrieval strategy: prioritizing recently published, directly responsive sources over pages with heavier accumulated editorial authority. Instead of leaning on a domain's long-term reputation, Perplexity's index favors whichever page answers the immediate query most precisely, updated continuously as new pages get published or existing ones get revised.

• Reference Rate

  Reference rate is a term some GEO platforms use as a synonym for citation rate, the percentage of relevant AI answers that reference a brand as a source.

• Retrievability

  Retrievability is the likelihood that a given piece of content gets surfaced at all during an AI system's initial document search, the step before any citation or answer-inclusion decision is made, meaning content can be well-written and still have low retrievability.

• Retrieval-Augmented Generation (RAG)

  Retrieval-Augmented Generation (RAG) is the architecture behind most AI search: the system retrieves a set of relevant documents or passages at query time, then generates its answer grounded in that retrieved material, rather than purely from what it learned during training.

• Schema Markup / Structured Data

  Schema markup, or structured data, is machine-readable metadata embedded in a page, most commonly as JSON-LD, that explicitly tells crawlers what the content actually is: an article, an organization, a product, a set of questions and answers, rather than leaving that classification to inference.

• Search-Grounded Retrieval

  Search-grounded retrieval is Gemini's mechanism of connecting a generated answer back to Google's own search index before applying a further citation filter on top. Rather than retrieving from an independent web crawl, Gemini effectively inherits whatever Google Search itself already considers relevant and trustworthy for a query, then layers its own criteria on that base.

• Semantic Search

  Semantic search is retrieval based on the meaning and intent behind a query, rather than exact keyword matching. It uses vector representations of text to find content that's conceptually related even when the actual wording is completely different.

• Sentiment

  Sentiment, in a GEO context, is the tone, positive, neutral, or negative, of how an AI system characterizes a brand whenever it mentions or cites that brand in a generated answer.

• Share of Model

  Share of model measures the number of mentions a brand receives from one or more specific large language models, expressed as a proportion of total brand mentions across the whole category, for a given set of queries.

• Speakable Schema

  Speakable schema is a structured-data property that marks specific passages within a page as suited for text-to-speech and voice-answer synthesis, originally built for voice assistants like Google Assistant, and increasingly relevant as conversational AI answers move toward spoken as well as written delivery.

• Statistical Density

  Statistical density is the number of concrete, quantified data points, figures, percentages, dates, named studies, per unit of content, functioning as a directly measurable proxy for how specific and citable a passage is rather than a subjective quality judgment.

• Statistics Addition

  Statistics addition is the Princeton GEO paper's tested tactic of replacing vague claims with specific numbers, sources, and dates, which the study measured as its single strongest-performing tactic, producing up to roughly a 41 percent improvement on the paper's PAWC visibility metric.

• Subjective Impression

  Subjective impression is the Princeton GEO paper's human-evaluated score for how visible and prominent a source feels within an AI-generated answer, collected by having human raters read the answer and judge a source's presence directly, as a complement to the automated PAWC metric.

• Topical Authority

  Topical authority is how comprehensively a site covers a subject area, measured across content depth, breadth, and internal linking between related pages, rather than any single page's individual quality, making it a property of a site's overall content structure.

• Zero-Click Search

  A zero-click search is a query fully answered inside an AI-generated response, with no need for the user to click through to any website. It's the outcome that makes citation, not ranking, the metric that actually determines whether a brand gets seen at all.

• knowsAbout

  knowsAbout is a schema property that explicitly declares a brand, organization, or author's areas of expertise or topical focus, giving AI systems a direct signal for topical relevance rather than requiring them to infer it purely from a site's published content.

• llms-full.txt

  llms-full.txt is an expanded version of llms.txt that includes a site's full page content, rather than short summaries, in one consolidated markdown file intended for AI systems to ingest directly. It trades the compact, curated nature of llms.txt for completeness, at the cost of a much larger file.

• llms.txt

  llms.txt is a markdown file placed at a website's root that summarizes its content and structure specifically for AI systems and their crawlers, functioning as the closest thing this space has to an emerging technical standard. It gives an AI system a curated map of a site instead of forcing it to infer structure from a full crawl.

• robots.txt (AI directives)

  robots.txt is the long-standing standard crawler-control file at a site's root, now extended with AI-specific user-agent rules that separately allow or disallow crawlers like GPTBot, ClaudeBot, and PerplexityBot. A single robots.txt file can permit standard search crawling while blocking one or more AI crawlers, or the reverse.

• sameAs

  sameAs is a schema property that links a brand or person's identity on one page to their authoritative external profiles elsewhere, commonly Wikidata, Wikipedia, and LinkedIn, giving crawlers a direct, machine-readable way to confirm that different mentions across the web refer to the same entity.