AI SEO Glossary (AEO / GEO Terms)

The vocabulary of AI SEO is unsettled. Different vendors use different labels for the same concept, the discipline borrows terms from information retrieval and machine learning research, and the MERIT Framework introduces chapter names that need their own definitions. This glossary collects the working definitions used throughout the Playbook so terms mean the same thing wherever they appear.

Where a term maps directly to a MERIT chapter, the entry links to the relevant chapter. Definitions are written for marketing leaders, not researchers, and lean toward operational meaning over academic precision.

Jump to: A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | R | S | T | V | Y | Z

A

AEO (Answer Engine Optimization)

The practice of optimizing to be the cited answer when AI systems produce direct answers rather than ranked lists of links. AEO is sometimes expanded as AI Engine Optimization. In the MERIT Framework, AEO is the cited-answer component of GEO (GEO equals AEO plus Brand SEO) and sits under the AI SEO umbrella. It is not a synonym for GEO: AEO earns the citation, while GEO also earns the recommendation.

AI Mode

Google's conversational AI search interface, distinct from AI Overviews. AI Mode delivers a chat-style experience with multi-turn dialogue, deeper synthesis, and more aggressive query fan-out. SE Ranking documented 9.2% URL consistency in AI Mode in August 2025, indicating substantially higher volatility than traditional Google search.

AI Overviews

Google's synthesized answer panels that appear above traditional search results. AI Overviews summarize information from multiple sources and link to a small set of cited pages. seoClarity research from February 2025 found that 97% of AI Overviews cite at least one source from the top 20 organic results, anchoring their selection in traditional ranking signals.

AI SEO

The umbrella term used throughout the MERIT Framework for the discipline of earning visibility in AI engines: search optimization evolved for AI retrieval. AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) sit under it as component disciplines, and LLM SEO is another umbrella-level label for the same work. AI SEO is the parent label because it does not pick a specific engine and applies uniformly across ChatGPT, Claude, Perplexity, Google AI Overviews, and Gemini. The MERIT Framework treats it as an evolution of SEO, not a separate discipline: Google's generative AI features are rooted in its core Search ranking and quality systems, so the same crawlable, authoritative, genuinely helpful content wins both. What changes is the depth of execution AI retrieval demands.

Anchor Set

A small, fixed group of 10 to 15 stable canary pages used to detect embedding or retrieval-model change before re-scoring an entire corpus. When the Anchor Set's similarity math moves without any content edits, the shift signals a model-side change rather than a content problem, which tells you to recompute scores corpus-wide. It is the early-warning instrument that keeps Information Gain Score programs from chasing false drift. See The Anchor Set.

Answer-First Content Architecture (Chapter 7)

The Relevance pillar chapter focused on structuring owned content for AI retrieval. Answer-First Content Architecture is about how content is presented (passage-level structure, snippet extractability, intent alignment; retrieval systems internally segment content, which is sometimes called chunking), not what content is created. See the Answer-First Content Architecture chapter.

B

Bingbot

Microsoft Bing's web crawler. Bingbot powers Bing search and is the primary content source for Microsoft Copilot. Treating Bingbot well is essential for any AI strategy involving Copilot or Bing-distributed answers.

BM25

A lexical retrieval scoring function that ranks documents by term frequency and inverse document frequency. BM25 remains a foundational scoring algorithm in modern hybrid retrieval systems, sitting alongside vector similarity to combine keyword precision with semantic recall. AI systems that use retrieval-augmented generation often blend BM25 with embedding-based scoring.

Brand SEO

The reputation, review, corroboration, and entity-recognition work that earns a brand a recommendation in AI answers, not merely a citation. In the MERIT Framework, Brand SEO is the half of GEO that AEO does not cover (GEO equals AEO plus Brand SEO); it maps to the Mentions pillar plus the narrative and reputation work in Transformation. Its inputs cannot be faked: real customer experience, real reviews, and genuine information gain.

C

ChatGPT

OpenAI's conversational AI product, the most prominent consumer-facing LLM and a primary surface for AI SEO. ChatGPT pulls from a combination of training data and real-time web search via OAI-SearchBot. Reddit citation rates in ChatGPT have ranged from 1% to 14% (Profound, October 2025).

Citation

An inline reference in an AI response that attributes information to a specific source. Citations in AI responses are the AI Search equivalent of organic search rankings. The goal of MERIT is to be the source AI cites when answering questions in your category.

Citation Half-Life

The median time it takes for an asset's AI-citation rate to fall by half. A short half-life means an asset decays quickly and needs frequent refreshing; a long half-life means it holds citation value with less maintenance. Measuring Citation Half-Life per asset type lets you set a refresh cadence on evidence rather than guess at it. See Citation Half-Life.

Citation rate

The frequency with which a brand, page, or source is referenced in AI responses to a defined set of queries. Citation rate is one of the four primary measurement metrics for AI SEO, alongside share of voice, sentiment, and AI-attributed traffic. As a working benchmark, citation rate above 25% on a tracked prompt set indicates category leadership.

ClaudeBot

Anthropic's web crawler. ClaudeBot is the user agent that Anthropic uses to fetch pages for retrieval and (per Anthropic's policy) for training Claude. Allowing ClaudeBot in robots.txt is part of the Crawler Access chapter (Chapter 11).

Claude

Anthropic's family of conversational AI models. Claude is widely deployed in enterprise contexts and consumer applications. Claude pulls from training data and, in some configurations, retrieval-augmented sources. Anthropic's bot identifier is ClaudeBot.

Co-citation

The pattern of being referenced alongside other authoritative sources within a single AI response or across a tracked prompt set. Co-citation drives AI visibility more reliably than format or rigor of the original asset alone. AI systems weight sources that appear together in trusted contexts as mutually reinforcing.

Common Crawl

A nonprofit foundation that publishes a free, open repository of web crawl data covering hundreds of billions of pages. Common Crawl is one of the foundational datasets used to train large language models. Being well represented in Common Crawl matters for training-baked knowledge of your brand.

Community Mentions and Positive Sentiment (Chapter 2)

The Mentions pillar chapter covering authentic community engagement on Reddit, Quora, LinkedIn, and industry forums. Build trust before promotion. Reach 500+ Reddit karma before any brand mention. Apply the 90/10 rule: 90% value, 10% promotion. See the Community Mentions and Positive Sentiment chapter.

Cosine similarity

A mathematical measure of similarity between two vectors, used in embedding-based retrieval to identify which passages are most relevant to a query. Cosine similarity scores between 0 and 1 describe how aligned two pieces of content are in semantic vector space. Higher cosine similarity means a passage is more likely to be retrieved for a given query.

Corpus Drift

Landscape-side decay: an unchanged page's Information Gain Score falls over time as competitors publish, entities shift, and category coverage expands around it. The page did not get worse; the surrounding corpus got better, so the same content now adds less net-new information. Detecting Corpus Drift is what triggers a refresh before citation rate erodes. See Corpus Drift.

Corpus Engineering

The systems-level discipline of designing, structuring, expanding, maintaining, and optimizing a body of content for AI retrieval and citation. Corpus Engineering treats a site's content as an engineered system with six components rather than a collection of individual pages. It is the broader practice within which Relevance Engineering sits as one component. See Corpus Engineering.

Corpus Engineering vs Relevance Engineering

The distinction between the whole and one of its parts. Relevance Engineering is the retrieval-optimization component focused on making individual passages findable and reusable; Corpus Engineering is the wider discipline that also covers corpus design, expansion, and maintenance across the full body of content. Relevance Engineering lives inside Corpus Engineering, not alongside it. See Corpus Engineering vs Relevance Engineering.

Crawler

An automated program that fetches web pages on behalf of a search engine or AI system. Major AI-relevant crawlers include GPTBot (OpenAI training), OAI-SearchBot (ChatGPT real-time search), ClaudeBot (Anthropic), PerplexityBot (Perplexity), Google-Extended (Google's AI training opt-out flag), and Bingbot (Microsoft and Copilot). Crawler configuration is the focus of the Crawler Access chapter.

Crawler Access (Chapter 11)

The Inclusion pillar chapter covering robots.txt, server-side rules, and access policy for AI crawlers. Crawler Access is foundational because no other chapter works if AI systems cannot fetch your content. See the Crawler Access chapter.

D

Disambiguation (entity)

The process by which AI systems determine which specific entity is being referenced when a name is ambiguous. Multiple businesses, people, and concepts can share names. Schema.org sameAs properties, consistent NAP data, and authoritative third-party references help AI systems disambiguate correctly.

E

E (Evidence)

The second pillar of MERIT. Evidence covers original, quantifiable assets that establish your brand as a primary source AI systems can reference and attribute. The Evidence pillar contains Original Source Asset Development, Information Gain Architecture, and Citation Reinforcement (Chapters 4 to 6).

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness)

Google's Quality Rater Guidelines framework for evaluating page quality. E-E-A-T is technically a traditional SEO concept, but with retrieval-augmented generation the underlying signals (author credentials, demonstrated experience, authoritative citation patterns) are critical inputs to AI citation. Treat E-E-A-T as foundational regardless of whether you call it traditional SEO or AI SEO work.

Embedding

A high-dimensional vector representation of text that captures semantic meaning. Embeddings allow AI systems to compute similarity between a query and content across a corpus, enabling retrieval of conceptually related passages even when keywords do not match. Modern AI Search relies heavily on embedding-based retrieval as the first stage of finding citation candidates. When an edit moves a page's embedding away from the queries it used to match, the result is a Vector Shift.

Embedding Audit

A cross-site workflow that computes Information Gain Score by embedding a target page and the competing top-of-SERP set, then measuring how far the target sits from that cluster. Run in Screaming Frog, it quantifies Embedding Strength: how distinct a page's vector position is relative to the pages it competes against. The Embedding Audit turns information gain from an editorial judgment into a measured number. See Embedding Audit.

Entity

A named thing (person, organization, product, place, concept) that AI systems represent as a node in their internal knowledge representation. Brands, founders, products, and topical domains are all entities. Entity-level recognition is more durable than keyword-level optimization because entities persist across phrasings.

Entity density

The concentration of clearly defined, well-disambiguated entities within a piece of content. Higher entity density (within reason) signals topical depth and helps AI systems classify and retrieve content more accurately. Entity density should be coherent with the page's topic, not stuffed.

Entity disambiguation

See Disambiguation. The deliberate work of telling AI systems which specific entity a name refers to using sameAs properties, consistent identifiers, and corroborating third-party references.

Entity Optimization (Chapter 10)

The Inclusion pillar chapter covering brand, people, product, and topical entity work. Entity Optimization is multi-faceted: it spans the organization itself, the founders and SMEs associated with it, the products and services it sells, and the topical domains it claims authority in. Build deep on core topics before going wide. See the Entity Optimization chapter.

F

Fan-out (query fan-out)

The technique AI systems use to decompose a single user query into multiple sub-queries, retrieve results for each, then synthesize a combined answer. Query fan-out is why AI Mode is more volatile than traditional search: each fan-out sub-query may pull a different set of sources. Optimizing for fan-out means thinking about the family of related questions a query implies, not just the literal phrasing.

Freshness

How recently a page was updated or published. AirOps research from March 2026 found that recently updated content earns a substantial citation lift over older content, and that most cited pages had been refreshed rather than left stale. Freshness compounds: stale content not only loses ranking, it loses citation eligibility. The slow decay of a corpus into staleness relative to the current retrieval landscape is Corpus Drift.

G

Generative Engine Optimization (GEO)

The practice of optimizing content and brand presence to be cited and recommended in generative AI responses. In the MERIT Framework, GEO equals AEO plus Brand SEO: the cited-answer work (AEO) plus the reputation, review, and entity work (Brand SEO) that earns a recommendation. GEO sits under the AI SEO umbrella alongside AEO. The term was popularized by academic research in 2023 and 2024.

Gemini

Google's family of multimodal AI models, deployed in Google's AI Overviews, AI Mode, and standalone Gemini products. Gemini draws from Google's index, Knowledge Graph, and trained capabilities.

Google-Extended

Google's user agent token used to control whether content is included in Bard, Gemini, and other Google generative AI training. Google-Extended is set independently of Googlebot, allowing publishers to remain in Google search while opting out of generative AI training. Configured in robots.txt.

GPTBot

OpenAI's user agent for crawling content used in training future models. Distinct from OAI-SearchBot, which fetches content for real-time ChatGPT search responses. Allowing GPTBot helps your brand become part of base training; allowing OAI-SearchBot helps your brand surface in real-time chat citations.

H

Hallucination

An AI response that contains confidently stated information that is factually incorrect. Hallucinations are most damaging when they describe your brand, products, or executives in ways that are wrong. Reputation alignment work (part of Transformation) addresses the question of whether the underlying error is training-baked or retrieval-baked, since the fix differs.

I

I (Inclusion)

The fourth pillar of MERIT. Inclusion covers technical accessibility and semantic precision that enables AI crawlers to discover, understand, and correctly interpret your content and entities. The Inclusion pillar contains Entity Optimization (Chapter 10), Crawler Access (Chapter 11), and IndexNow (Chapter 12).

IndexNow

An open-source protocol for instantly notifying search engines when content is added, updated, or deleted. IndexNow is supported by Bing, Yandex, and a growing list of partners. Cloudflare integrates IndexNow at the CDN level. See the IndexNow chapter.

Information gain (net-new information gain)

The concept that AI systems reward content that adds something the model has not already seen across thousands of other sources. AirOps's March 2026 analysis confirms LLMs filter for sources that contribute new information rather than restating existing content. Information gain is the mechanism behind the Original Source Asset Development chapter (Chapter 4); original assets are the vehicle, information gain is what they deliver. Searchbloom measures it with two layers: Information Gain Density (the editorial count) and the geometric Information Gain Score.

Information Gain Density (IGD)

The editorial layer of information gain measurement: a count of the distinct, original, attributable insights a page contributes relative to the competitor saturation set. It is governed by the 5-to-7 Rule, which holds that a page needs roughly 5 to 7 genuinely net-new insights to read as a primary source rather than a restatement. IGD is the human-judged complement to the geometric Information Gain Score. See Information Gain Density.

Information Gain Score (IGS)

The geometric layer of information gain measurement: 1 minus the maximum cosine similarity between a page and the top-10 SERP set, expressed on a 13-grade letter scale. A high IGS means the page sits far from everything already ranking, so it contributes vector-distinct information a model has not seen. IGS is the measured counterpart to the editorial Information Gain Density. See Information Gain Score.

Intra-Site Embedding Audit

A pairwise cosine-similarity pass across a site's own pages to surface internal duplication, content gaps, and topical drift. Where the cross-site Embedding Audit measures distance from competitors, the Intra-Site Embedding Audit measures distance among your own pages so near-duplicates can be consolidated and thin coverage filled. It keeps a corpus internally distinct as it grows. See Intra-Site Embedding Audit.

J

JSON-LD

JavaScript Object Notation for Linked Data, the recommended format for embedding Schema.org structured data on web pages. JSON-LD is preferred because it does not interfere with rendered HTML and supports complex nested entity definitions. LLMs do not parse JSON-LD directly at generation time, but search engines do, and that traditional SEO layer feeds the retrieval signals AI systems rely on.

K

Knowledge Graph (Google Knowledge Graph)

Google's internal knowledge representation that stores entities and the relationships between them. Knowledge Graphs are not created by implementing schema. They are created by Google when entities reach sufficient prominence through demand and search volume. Schema accelerates entity disambiguation but does not generate Knowledge Panel coverage on its own.

Knowledge Panel

The visible card that appears in Google search results for prominent entities, drawn from the Knowledge Graph. When Google creates a Knowledge Panel for your brand, claim it immediately. Unclaimed panels can be edited by competitors, former employees, or other parties with their own agendas.

L

Latent intent

The underlying goal behind a search query that may not be stated explicitly. AI systems infer latent intent and answer the implied question, not just the literal phrasing. Optimizing for latent intent means understanding the family of related questions a query implies and addressing the actual buyer or researcher need.

LLM (Large Language Model)

A neural network trained on massive amounts of text capable of generating coherent natural language responses. LLMs power ChatGPT, Claude, Perplexity, Gemini, and the AI components of Google AI Overviews and Microsoft Copilot. LLMs combine information stored during training with retrieval from external sources.

LLM SEO

A label for AI SEO that emphasizes the LLM as the target. The MERIT Framework treats LLM SEO as another umbrella label for AI SEO, distinct from its component disciplines: AEO (being the cited answer) and GEO (which equals AEO plus Brand SEO).

M

M (Mentions)

The first pillar of MERIT. Mentions covers third-party validation across trusted platforms in your industry where AI systems discover authoritative signals about your brand, products, services, and expertise. The Mentions pillar contains Pay-to-Play Placements (Chapter 1), Community Mentions and Positive Sentiment (Chapter 2), and Third-Party Corroboration (Chapter 3).

Measurement Cadence and Expectations (Chapter 13)

The Transformation pillar chapter covering weekly monitoring, monthly dashboards, and quarterly strategic reviews. Measurement Cadence and Expectations specifies the operating rhythm that turns AI Search visibility into a managed program rather than a one-time project. See the Measurement Cadence and Expectations chapter.

MERIT Framework

A methodology for AI SEO developed by Cody C. Jensen of Searchbloom. MERIT consists of five pillars (Mentions, Evidence, Relevance, Inclusion, Transformation) containing fifteen chapters. The framework is designed for marketing leaders running AI Search programs at mid-market and enterprise scale.

Multimodal

Capable of processing or generating multiple input or output modalities (text, image, audio, video). Modern AI systems are increasingly multimodal, which expands the surfaces that matter for AI SEO beyond text pages. YouTube, podcast transcripts, and image alt text all become first-class inputs.

N

Sentiment Footprint and Sentiment Shaping (Chapter 14)

The Transformation pillar chapter covering whether the story AI systems tell about your brand matches the story you want told. Sentiment Footprint and Sentiment Shaping is what you check when AI responses about your business are technically accurate but strategically off. See the Sentiment Footprint and Sentiment Shaping chapter.

Net-new information gain

See Information gain. The mechanism behind original source asset development.

O

OAI-SearchBot

OpenAI's user agent for fetching content during real-time ChatGPT search. Distinct from GPTBot, which is used for training. Configuring access for OAI-SearchBot is more time-sensitive than GPTBot because it directly affects what ChatGPT can cite right now.

Organizational Evolution (Chapter 15)

The Transformation pillar chapter covering how marketing organizations need to evolve to execute AI Search well. Marketing is becoming engineering: systematic measurement, version-control thinking, prompt and system design, AI-collaborative production. See the Organizational Evolution chapter.

Original Source Asset Development (Chapter 4)

The Evidence pillar chapter covering the creation of frameworks, opinion, research, data, and tools that AI systems can cite as primary sources. The mechanism is information gain: AI rewards content that contributes something the model has not seen elsewhere. See the Original Source Asset Development chapter.

P

Passage retrieval

The retrieval pattern in which AI systems pull individual passages or sections from a page rather than the page as a whole. Passage retrieval is why passage-level structure, snippet extractability, and self-contained sections matter. A page might be cited only for one of its subsections; the rest is invisible to that response.

Passage-level structure

Dividing content into self-contained, semantically coherent passages sized for retrieval. A well-structured page can have multiple passages cited independently; a poorly structured page is either retrieved as a whole (losing precision) or skipped entirely. Passage-level structure is a primary technique in Answer-First Content Architecture.

Pay-to-Play Placements (Chapter 1)

The Mentions pillar chapter covering premium placements on trusted review and directory platforms (G2, Clutch, Capterra, Gartner, Trustpilot). AI systems cannot see who paid, but they read the higher rankings that paying produces. See the Pay-to-Play Placements chapter.

Perplexity

An AI search engine that produces conversational answers grounded in real-time web sources with prominent inline citations. Perplexity has the highest Reddit citation rate among major AI engines (6.3%, per Profound, October 2025). Perplexity tends to cite a wider variety of sources than ChatGPT or Google AI Overviews.

PerplexityBot

Perplexity's web crawler. Allowing PerplexityBot is part of the Crawler Access chapter (Chapter 11).

Personalization (in AI Search context)

The degree to which AI responses vary by user, account, geography, prior conversation history, or other personalization signals. Personalization adds another layer of volatility on top of platform inconsistency. Measurement methodology should account for it by sampling across accounts, sessions, and prompt phrasings.

Probability distribution (for visibility)

The honest framing for AI Search visibility. Rather than thinking of AI citation as deterministic ("we rank #1"), think of it as a probability distribution: across many runs of similar prompts, what fraction of responses cite us, and where do we sit relative to competitors? SparkToro's January 2026 research found AI brand recommendations are statistically random more than 99% of the time at the per-prompt level, which is why distributional thinking is required.

R

R (Relevance)

The third pillar of MERIT. Relevance covers comprehensive, intent-aligned content structured for AI retrieval in self-contained, citable segments. The Relevance pillar contains Answer-First Content Architecture, Multi-Format Surface Coverage, and Semantic HTML and Entity-Rich Language (Chapters 7 to 9).

RAG (Retrieval-Augmented Generation)

An architecture in which an LLM retrieves relevant external documents at query time and uses them to generate its response. RAG is the dominant pattern in modern AI Search. The retrieval layer determines which sources are eligible to be cited; the generation layer produces the actual answer text. Both layers matter, but the retrieval layer is where MERIT does most of its work.

Reddit

The community discussion platform that has become a dominant citation source for AI systems. Profound's October 2025 analysis ranked Reddit second only to YouTube as the most-cited source overall, with citation rates of 1.2% in ChatGPT, 2.3% in Google AI Overviews, and 6.3% in Perplexity. Reddit engagement is governed by the Community Mentions and Positive Sentiment chapter (Chapter 2).

Relevance Engineering

The practice of structuring content so AI systems can find, parse, and reuse it. Relevance Engineering is the operational discipline behind the Answer-First Content Architecture chapter (Chapter 7). It draws on information retrieval research, content design, and a working understanding of how AI retrieval models score candidate passages.

Reputation alignment

The work of bringing AI-generated descriptions of your brand into alignment with the truth. The first diagnostic question is whether an inaccuracy is training-baked (the model learned the wrong thing) or retrieval-baked (the model is reading current but inaccurate sources). The fix differs: retrieval-baked errors can be corrected by fixing the underlying source; training-baked errors require time and corroborating new signals.

S

sameAs (schema property)

A Schema.org property used to declare that an entity on your site is the same entity as one on another platform (LinkedIn, Wikipedia, Crunchbase, official directories). sameAs links are the primary mechanism for entity disambiguation. Implement sameAs across organization, person, and product schema wherever the entity has a verified presence on a third-party platform.

Schema.org

A collaborative vocabulary of structured data types maintained by Google, Microsoft, Yahoo, and Yandex. Schema.org markup helps search engines understand what a page is about and how its entities relate. LLMs do not parse schema directly at generation, but search engines do, and the resulting structured signals shape the retrieval layer that AI Search depends on.

Semantic triples

The subject-predicate-object structure used to represent facts in knowledge graphs. "Searchbloom is headquartered in South Jordan, Utah" is a semantic triple: subject (Searchbloom), predicate (is headquartered in), object (South Jordan, Utah). Writing clear, factual triples in your content helps AI systems extract structured knowledge.

Semantic-Relationship Drift

The slow reshuffling of entity adjacency and salience over time: which entities sit semantically near each other, and how strongly. A page can stay unchanged while the relationships AI systems infer around its entities move, so it is read in a different context than when it was published. Tracking Semantic-Relationship Drift catches narrative and entity decay that page-level metrics miss. See Semantic-Relationship Drift.

Setting Expectations

Executive education, KPI selection, realistic timelines, and addressing common misconceptions about AI Search. Volatility is real, randomness is real, and stakeholders need to understand both before measurement begins. This work is folded into the Measurement Cadence and Expectations chapter (Chapter 13). See the Measurement Cadence and Expectations chapter.

Share of voice

The percentage of AI responses on a tracked prompt set that cite your brand relative to total competitor citations. As a working benchmark, share of voice above 25% in your category indicates leadership, and below 10% indicates significant gap to category leaders.

Snippet extractability

The degree to which a passage can be lifted from a page and presented as a self-contained answer. Snippet extractability requires that a section makes sense on its own, contains the specific information promised by its heading, and does not depend on context elsewhere on the page.

Structured data

Information on a page formatted in a machine-readable way, typically using Schema.org vocabulary in JSON-LD. Structured data is a traditional SEO factor that indirectly supports AI Search by feeding the entity layer that retrieval depends on.

T

T (Transformation)

The fifth pillar of MERIT. Transformation covers systematic measurement and iterative optimization that accounts for AI platform volatility while driving sustained visibility improvements. The Transformation pillar contains Measurement Cadence and Expectations (Chapter 13), Sentiment Footprint and Sentiment Shaping (Chapter 14), and Organizational Evolution (Chapter 15).

Third-Party Corroboration (Chapter 3)

The Mentions pillar chapter covering guest posting, co-authorship, content syndication, and multi-source validation. AirOps research found that approximately 85% of AI citations come from third parties rather than the brand's own site. Third-Party Corroboration is the chapter that earns those citations. See the Third-Party Corroboration chapter.

Top-20 organic results

The first two pages of traditional Google search results. seoClarity research from February 2025 found that 97% of AI Overviews cite at least one source from the top 20 organic results. This is a primary reason AI Search is an evolution of SEO rather than a separate discipline: the eligible citation pool is anchored in the same ranking signal.

Transformer (architecture)

The neural network architecture introduced in the 2017 "Attention Is All You Need" paper that underpins virtually every modern LLM. Transformers process sequences using self-attention, which lets the model weight the importance of different tokens relative to each other. Marketing leaders do not need to understand the math, but the term appears often enough in vendor materials that it is worth knowing.

V

Vector Drift

Model-side change in which an embedding-model upgrade rotates the coordinate system itself, so the same content lands at different cosine distances without any edit. The page did not move; the math under it did, which can shift Information Gain Score corpus-wide overnight. Vector Drift is why an Anchor Set of stable canary pages is monitored before re-scoring. See Vector Drift.

Vector Shift

The change in a page's vector position relative to the SERP cluster after a content edit: whether the page moved toward or away from the pages it competes against. A favorable Vector Shift is the receipt that real information-gain work landed, distinguishing genuine differentiation that was earned from superficial changes that were gamed. It is the before-and-after measurement that proves an edit changed retrieval standing. See Vector Shift.

Y

YouTube (as AI citation surface)

The video platform that, per Profound's October 2025 analysis, is the most-cited source overall across major AI engines. YouTube's transcripts feed AI systems directly. Optimizing video titles, descriptions, captions, and chapter markers materially affects whether AI cites your video content.

Z

Zero-click search

A search interaction in which the user gets their answer from the SERP itself (or an AI Overview) without clicking through to any source page. Zero-click is the dominant mode for AI Overviews and a growing share of AI Mode. The implication is that traditional traffic metrics undercount your actual audience. AI-attributed visibility (citation rate, share of voice) becomes the leading indicator; downstream traffic and conversion become trailing indicators.