AI search & citation

173 terms in this topic.

• Actor certificate

  An actor certificate is the cryptographic identity a signing party uses to attach an action to a C2PA content credentials manifest — the thing that turns ‘someone said this asset is AI-generated’ into ‘this specific certified actor said it, and the signature verifies.’ Actors can be generators (an AI model provider certifies c2pa.ai_generated actions), editing tools (Photoshop, Canva, Adobe Express certify c2pa.edited), or brands (the brand certifies c2pa.published to attach publication to its public identity).

• Agent-readable site

  An agent-readable site is one built so that the second audience for a website — the AI crawlers and shopping/task agents that feed AI answers — can ingest it cleanly, in addition to human visitors. This audience does not render CSS, scroll, or click; it ingests.

• Agentic shopping

  Agentic shopping is the use of an AI agent — Amazon Rufus, ChatGPT, Perplexity, Google AI Mode — to research, compare, and execute a purchase on the user’s behalf, with progressively less manual interaction.

• AI citation rate

  AI citation rate is the share of relevant LLM-generated answers in which a brand or page is cited as a source — the AI-search equivalent of organic click-through rate. It's measured across major AI surfaces: ChatGPT, Perplexity, Claude, Google AI Overviews, Google AI Mode, Bing Copilot, Amazon Rufus.

• AI content marketing

  AI content marketing is the use of generative models within a content programme — drafting copy, producing imagery, generating variants, or assisting research and briefing. It is best understood as a change in the cost curve rather than a change in what content marketing is: the marginal cost of an additional asset falls sharply, while the cost of deciding what is worth making does not move at all.

• AI crawler

  An AI crawler is an automated agent that fetches web content to feed an AI system — training data pipelines, retrieval-augmented answer engines, and shopping/task agents — rather than to build a classic search index.

• AI grounding

  AI grounding is the practice of anchoring an AI model's response in verifiable, retrievable source material — typically via retrieval-augmented generation (RAG), search-tool calls, or a structured knowledge base — so the answer cites real sources rather than fabricating facts.

• AI Overviews

  AI Overviews are the AI-generated answer summaries Google places above traditional search results, drawing from indexed pages, structured data, and high-authority sources to synthesize a short answer for the query. AI Overviews evolved from Search Generative Experience (SGE) and are now the default for most informational queries in the US, UK, India, Japan, and an expanding list of markets.

• AI search attribution

  AI search attribution is the discipline of connecting AI-engine citations and assistant recommendations to downstream revenue — sessions, conversions, and pipeline — inside an analytics stack that was originally built for blue-link referrers.

• AI search citation

  An AI search citation is a reference to a webpage, brand, or piece of content inside an answer produced by a generative AI search engine — ChatGPT, Perplexity, Google AI Overviews, Microsoft Copilot, Claude.

• AI search visibility

  AI search visibility — also called AI Visibility, AEO (Answer Engine Optimization), or GEO (Generative Engine Optimization) — is the discipline of optimizing content to be cited, quoted, or referenced inside the answers generated by AI search engines and assistants: Google AI Overviews, Perplexity, ChatGPT Search, Claude, Microsoft Copilot, Amazon Rufus, and TikTok Symphony.

• AI shopping assistant

  An AI shopping assistant is a generative-AI product surface that helps shoppers find, compare, and decide on purchases through natural-language conversation rather than keyword search and filter clicks. The major players as of 2025: Amazon Rufus (embedded in Amazon), Perplexity Shopping (with Shopify integration), ChatGPT Shopping (rolled out in 2024–2025), Google's Shopping AI Mode, and Microsoft Copilot's product surfaces.

• AI shopping feed

  An AI shopping feed is the product data feed the AI shopping assistant layer (ChatGPT Shopping, Perplexity Pro Shop, Gemini Commerce, Copilot Commerce, Amazon Rufus surface) consumes for semantic matching, ranking, and citation.

• AI Shopping Optimization (AISO)

  AI Shopping Optimization (AISO) is the practice of structuring product content, reviews, schema, and brand entities so a product is preferentially surfaced inside AI shopping assistants — Amazon Rufus, Perplexity Shopping, ChatGPT Shopping, Google AI Mode shopping panels, and Microsoft Copilot.

• AI visibility tracking

  AI visibility tracking is the discipline of measuring a brand's presence across AI search and shopping surfaces — ChatGPT Search, Perplexity, Google AI Mode, Microsoft Copilot, Amazon Rufus, Claude, and the agentic-shopping checkout flows landing in late 2026 — and converting the measurements into an actionable content backlog.

• AI watermark

  An AI watermark is a signal embedded in AI-generated content — visible or invisible — that identifies it as synthetic. Visible watermarks (a logo or label burned into the corner) are the legacy form, mostly replaced by invisible techniques because visible marks reduce commercial usability.

• Amazon Rufus

  Amazon Rufus is Amazon's generative AI shopping assistant, launched broadly in 2024 and embedded directly in the Amazon app and website. Rufus answers shopper questions in natural language ('what's the best running shoe for flat feet under $100?', 'compare these two air fryers', 'is this safe to use on hardwood?') by synthesizing across product listings, customer reviews, Q&A sections, and external web content.

• Anchor drift

  Anchor drift is the shift in which sentence inside a rerank-surviving chunk wins the anchor slot across sessions, even when the chunk itself has not been edited. Mid-2026 cohort: roughly 28% of anchor sentences on rerank-surviving chunks shift within a rolling 8-week window across the general-purpose engines, with Perplexity running the highest drift rate (roughly 38% within 8 weeks) and the other engines running 18–24%.

• Anchor signal density

  Anchor signal density is the sentence-level property that measures how much citable signal (numeric statistics, named-entity tokens, head-query keywords) sits in the first 40% of a candidate anchor sentence. Sentences with both a numeric statistic and a named-entity token front-loaded in the first 40% win the anchor slot at 2.1× the rate of equivalent sentences with either signal alone or with the signals back-loaded past the sentence's midpoint.

• Anchor token load

  Anchor token load is the sentence-length property the anchor picker scores against the engine's preferred anchor length range — 17–28 words on desktop and 12–22 words on mobile for the general-purpose engines in mid-2026. Sentences in the preferred range win the anchor slot at 1.4–1.8× the rate of sentences outside the range.

• Anchor-slot survival rate

  Anchor-slot survival rate is the headline metric of a sentence-level anchor-engineering program: anchor slots won divided by verbatim citations available on the same rerank-surviving chunk universe, scored per priority page on a rolling 4-week window. A rate above 68% is category-leading; 50–68% is competitive; below 50% is exposed.

• Answer engine optimization (AEO)

  Answer engine optimization (AEO) is the discipline of structuring content so it surfaces and gets quoted directly inside answer-first interfaces — Google AI Overviews, ChatGPT, Perplexity, Microsoft Copilot, Amazon Rufus, voice assistants, and rich snippets.

• Answer inclusion rate

  Answer inclusion rate is the share of a target set of comparison or consideration queries for which a brand actually appears in the AI-generated shortlist. It is distinct from citation rate, which measures single-source factual citations: inclusion rate measures presence in assembled 'best X' and 'X vs Y' answers, where being in the list at all is the first hurdle.

• Answer-anchor sentence

  The answer-anchor sentence is the single sentence-length hyperlink text an AI search engine renders as the visible citation on the answer surface. Every major engine through 2026 renders a verbatim citation as three surfaces — a numbered source chip, a short quoted or bolded fragment, and a sentence-length hyperlink anchor the user actually clicks.

• Assistant handoff rate

  Assistant handoff rate is the percentage of AI-assistant recommendations on a defined query set that result in a click-through to the brand’s own surface — the PDP, the homepage, or the comparison page — vs. completing inside the assistant (add-to-cart, assistant-driven order, or no action).

• Assistant-side revenue

  Assistant-side revenue is the slice of brand revenue captured inside an AI assistant — via in-flow checkout, in-assistant subscription start, or assistant-driven add-to-cart that completes without ever generating a session on the brand’s own analytics stack.

• Author entity

  An author entity is the resolvable, credentialed identity of the person credited with a piece of content — expressed through structured author schema (Person markup), a consistent bio, a sameAs graph of the author's own profiles (LinkedIn, a personal site, ORCID where relevant), and a visible credential trail.

• Author-authority audit

  An author-authority audit is the systematic review of a site’s credibility layer for AI search — checking, page by page, whether the bylines resolve to real author entities, whether the claimed expertise is externally corroborated, whether the content carries first-hand experience, and whether the body has the expertise density its byline promises — run before adding more content on top of an unaccountable foundation.

• Brand entity graph

  A brand entity graph is the connected set of structured facts an AI engine holds about a brand — who founded it, what category it sits in, which products it sells, which competitors it is associated with, which use cases customers describe in reviews, which media sources cover it, and which authoritative pages link to its own site.

• Brand hallucination

  A brand hallucination is an AI engine stating a wrong fact about a specific brand with total confidence — a discontinued feature described as current, stale pricing, a wrong founder, a nonexistent partnership, or facts belonging to a similarly-named competitor.

• Brand mention monitoring

  Brand mention monitoring, in an AI-search context, is the practice of systematically querying AI engines and scanning the off-site corpus to track how a brand is mentioned, cited, recommended, and factually described across ChatGPT, Perplexity, Google AI Mode, Copilot, and Amazon Rufus.

• C2PA (content provenance)

  C2PA (Coalition for Content Provenance and Authenticity) is the open technical standard for cryptographically signed content provenance — a way to attach tamper-evident metadata to images and video that records what created the asset (camera, AI model, editing software), when, and how it was modified after capture.

• Candidate set

  The candidate set is the chunk-level shortlist an AI search engine's retrieval stage returns to the rerank stage per sub-query — typically 40–120 chunks on Google AI Mode and Microsoft Copilot, 60–90 on ChatGPT Search, 40–80 on Perplexity, and 30–60 on Claude in mid-2026.

• Card thumbnail slot compliance

  Card thumbnail slot compliance is the citation-card sub-property that measures whether the AI search engine renders a thumbnail image on the citation card at the correct card-preview ratio. Pages exposing at least one card-preview-ratio image (typically 1.91:1 open-graph ratio sized 1200×628 or higher) render a thumbnail that lifts card-surface click-through 1.5× on Perplexity and 1.2× on Google AI Mode.

• Card timestamp signal

  Card timestamp signal is the citation-card sub-property that measures whether the AI search engine renders a timestamp chip on the citation card stamped with a date the freshness pipeline agrees to trust.

• Card title truncation compliance

  Card title truncation compliance is the citation-card sub-property that measures whether the page's title tag renders within the AI search engine's card-title character budget without truncating a load-bearing token. Page titles that render within budget (44–70 characters depending on engine, with mobile budgets running 20–30% shorter than desktop) capture card-surface click-through 1.3–1.5× over titles that truncate mid-word or mid-brand-name.

• Category-defining query

  A category-defining query is the small set of high-volume, high-intent queries an AI engine treats as canonical for a product category — the queries that resolve to a 1–5 shortlist recommendation, capture the bulk of agentic-shopping revenue, and crystallize 9–18 months after the assistant launches in a category.

• ChatGPT Search

  ChatGPT Search is OpenAI's integrated web-search experience inside ChatGPT — launched October 2024, generally available to free users by 2025. Instead of routing queries to a separate search engine, the user asks ChatGPT a question, the model performs live web search behind the scenes, then composes a sourced answer with inline citation links.

• ChatGPT Shopping

  ChatGPT Shopping is OpenAI's product-discovery surface inside ChatGPT, rolled out across 2024–2025. When a user asks a buying question ('best noise-cancelling headphones under $300 for a frequent flyer'), ChatGPT returns a structured set of product recommendations with images, prices, key specs, and merchant links — sourced from a mix of retailer feeds, product reviews, third-party comparison content, and editorial sources.

• Chunk overlap window

  The chunk overlap window is the small character buffer (~80–100 characters on the major mid-2026 substrates) that overlaps between adjacent chunks. The overlap carries cross-chunk context the embedding alone cannot capture — when a sentence late in chunk N references an entity introduced early in chunk N, the overlap preserves the reference and the chunk embedding holds resolution against the entity.

• Chunk retrieval

  Chunk retrieval is the step in an AI search pipeline where a page is split into passages and individual passages, rather than whole documents, are scored and selected as candidates for an answer.

• Chunk-rationale alignment

  Chunk-rationale alignment is the structural property of a heading-bounded passage where the chunk's closing synthesis sentence is the rationale snippet the engine surfaces alongside the citation. The two artifacts converge: a well-formed chunk closes with one synthesis sentence that resolves the chunk's claim, and that same sentence is the rationale the substrate lifts.

• Citable claim shape

  A citable claim shape is the sentence-level structure the synthesis stage rewards with verbatim citation — a leading sentence that names the entity, asserts the claim, and quantifies the assertion in a self-contained span the engine can lift into the rendered answer without surrounding context.

• Citation card drift

  Citation card drift is the shift in citation-card composition — publisher badge, favicon rendering, page-title truncation, timestamp chip, thumbnail, or card position — on an anchor-slot-winning chunk across sessions, even when the underlying page has not been edited. Mid-2026 cohort: roughly 24% of citation cards on anchor-slot-winning chunks shift within a rolling 8-week window across the general-purpose engines.

• Citation card position weight

  Citation card position weight is the citation-card sub-property that measures where the card renders in the AI search engine's source strip (position #1 through position #4–10 depending on engine). Cards rendered in the top-3 source slots capture 74–88% of card-surface click-through depending on the engine — Copilot 86% (source #1–3), Google AI Mode 84%, ChatGPT Search 78%, Perplexity 74%, Claude 88% (source #1–2 alone).

• Citation card rendering

  Citation card rendering is the post-anchor UI-layer surface AI search engines compose alongside the anchor sentence — favicon, publisher-badge string, truncated page-title fragment, timestamp chip, thumbnail slot, and card position in the rendered source strip.

• Citation drift

  Citation drift is the week-over-week delta in citation share on a single query, single engine — the early-warning indicator that a competitor has published into your space, that an engine has re-weighted its retrieval substrate, or that one of your pages has decayed below the freshness threshold the engine prefers.

• Citation footprint

  A citation footprint is the full inventory, for a single brand, of every URL the major AI engines cite when that brand is the cited source — paired with the rationale snippet the engine attached, the query that triggered it, the engine, and the week the run happened on.

• Citation share

  Citation share is the percentage of AI-engine citations on a defined query set that name your brand, normalized by the total citations the engine surfaced across that query set. It is the AI-search analogue of share-of-voice in classic SEO, and the only durable headline metric for an AI-visibility program.

• Citation slot weight

  Citation slot weight is the per-position multiplier the synthesis stage applies to surviving reranked chunks when composing the rendered answer. Mid-2026 anchors across the major engines: slot 1 carries roughly 1.0 weight (the answer's opening sentence routes its rendered claim from this chunk on 60–80% of answers); slot 2 carries 0.55–0.7; slot 3 carries 0.3–0.45; slots 4–8 carry 0.1–0.2 combined.

• Citation-vs-paraphrase decision

  The citation-vs-paraphrase decision is the per-chunk step the synthesis stage runs over every surviving reranked candidate — does this chunk get rendered as a verbatim quote in the answer (with a numbered source chip and quoted span the user reads as the engine's evidence), or as a paraphrased composition in the engine's voice (with a numbered source chip but no quoted span), or dropped from the rendered answer entirely (source relegated to the 'further sources' panel).

• Comparison answer

  A comparison answer is the output an AI search engine generates for a 'best X', 'top 5', 'X vs Y', or 'alternatives to X' query — a shortlist, table, or ranked prose the model assembles rather than retrieves.

• Consensus signal

  The consensus signal is the aggregate of what the open web says about a brand — across community discussion (Reddit, forums, Q&A sites), review aggregates, and third-party editorial (‘best X’ listicles, independent comparisons) — that an AI engine reads to decide whether to recommend that brand in a commercial answer.

• Consideration set

  In AI search, the consideration set is the shortlist of candidate options an answer engine assembles before it ranks anything for a comparison query. It is the decisive, most-overlooked stage of a comparison answer: a brand that never enters the set cannot be ranked, characterized, or shown, no matter how strong its individual pages are.

• Content decay

  Content decay is the gradual loss of search ranking, traffic, and citation strength a piece of content experiences over time, even when nothing about the content itself changes. Decay happens because the surrounding ranking environment evolves — competitors publish fresher coverage, the topic's entity graph shifts, freshness signals depreciate, and search-rater feedback re-trains ranking models.

• Content decay curve

  The content decay curve is the measurable trajectory citation share rides on an un-refreshed priority page as it ages past the engine's freshness window. The curve has three phases: latency (months 0–6, share holds or trends mildly up as inbound links accumulate), compression (months 6–12, share compresses 15–35% relative to peak as the substrate begins weighting age into the retrieval ranking and the page loses long-tail share first), and drift (months 12+, share drifts toward zero on commercial queries at 8–12% per month and recovery requires a full content refresh rather than a cosmetic date update).

• Content gap (AI search)

  A content gap in the AI-search context is a (URL template, query cluster) pair where competitors hold consistent shortlist position and the brand holds none — measured against a stable priority query set across the major AI engines (ChatGPT Search, Perplexity, Google AI Mode, Microsoft Copilot, Amazon Rufus, Claude).

• Content gap analysis

  Content gap analysis is the process of identifying topics, keywords, questions, and content types that your target audience searches for but your website does not adequately cover. The goal is to find opportunities where creating new content (or improving existing content) can capture search traffic, answer user questions, and fill gaps in the buyer journey.

• Conversation state carryover

  Conversation state carryover is the signal the engine carries forward from earlier turns into the retrieval, rerank, and synthesis stages of the follow-up turn. Through 2026 every major engine carries at least four state signals across turns: the head-turn cited URLs (which bias retrieval toward the same source set on the follow-up), the head-turn rationale snippets (which bias rerank toward chunks with similar claim shapes), the head-turn entity slot (which biases the synthesis stage toward the same brand-entity disambiguation), and the head-turn user reaction signals (which bias the entire pipeline away from sources the user appeared to reject).

• Conversation thread retention

  Conversation thread retention is the per-engine memory window inside a single AI search session — the number of prior turns the engine carries forward into the retrieval, rerank, and synthesis stages of the next turn.

• Conversational follow-up turn

  The conversational follow-up turn is the user's second (and subsequent) message in an AI search conversation — the message that follows the engine's first answer with a refinement, a comparison ask, a clarification, or a pivot to a related sub-topic.

• Cosmetic refresh discount

  The cosmetic refresh discount is the penalty the AI substrate applies when a page's visible last-updated date or schema dateModified moves forward while the content-diff hash (the substrate's hash of the page text between successive crawls) shows no meaningful prose change.

• Credential graph

  A credential graph is the machine-readable web of links from an author entity to the external profiles and records that corroborate the author’s claimed expertise — sameAs links to LinkedIn, a personal site, X, ORCID or a professional registry, published work, and speaking or media appearances — that lets an AI engine confirm the person is real and the credentials are genuine.

• Cross-encoder rerank

  Cross-encoder rerank is the specific model architecture every major AI search engine runs through 2026 for its post-retrieval rerank pass. Unlike the bi-encoder used at retrieval (which encodes the query and each chunk independently into dense vectors and scores by similarity), a cross-encoder reads the (sub-query, chunk) pair jointly through the transformer attention layers and outputs a single relevance score that captures the interaction between the two strings.

• E-E-A-T

  E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trustworthiness — Google's framework for evaluating content quality, codified in its Search Quality Rater Guidelines. The first 'E' (Experience) was added in December 2022, joining the original E-A-T from 2014. E-E-A-T is not a direct ranking signal but a meta-framework that the rater workforce uses to score search results, which then trains Google's ranking models.

• Engine refresh cycle

  An engine refresh cycle is the cadence on which an AI search substrate re-crawls and re-embeds an already-indexed source against the latest fan-out queries, re-scoring its retrieval rank against fresher competitors.

• Entity disambiguation

  Entity disambiguation is the AI-engine process of deciding which specific brand, product, person, or place a query refers to when the surface words are ambiguous. ‘Acme’ could be the cleaning brand, the SaaS company, or a fictional reference; ‘Pulse’ could be a watch brand or a fitness app.

• Entity home

  An entity home is the single canonical URL a brand designates as the authoritative definition of its entity — usually the homepage or a dedicated about/company page. It carries the authoritative Organization schema, the full sameAs graph, and the canonical spelling, description, founding facts, and category framing that every other page and third-party profile should agree with.

• Entity SEO

  Entity SEO is the practice of optimizing content for the entities (people, places, products, organizations, concepts) that search engines and LLMs use to organize knowledge, rather than for raw keyword strings. The technical foundation is the knowledge graph — a structured database (Google's, Microsoft's, or Wikipedia's Wikidata) that maps entities and their relationships.

• Experience signal

  The experience signal is the evidence that a piece of content comes from genuine first-hand engagement with its subject rather than a rewrite of what already ranks — the operational form of the first ‘E’ (Experience) in E-E-A-T, and the credibility marker an AI engine most rewards because it is the one the engine cannot synthesize from other sources.

• Expertise density

  Expertise density is the concentration of accurate, specific, non-obvious substance per page — the depth an AI engine can read in the content itself, independent of who is credited with it.

• Fact-correction loop

  The fact-correction loop is the repeatable four-stage workflow for remediating wrong facts an AI engine states about a brand: detect, diagnose, correct, verify. Detect: run a brand-fact query set (‘how much does X cost’, ‘does X have feature Y’, ‘who founded X’, ‘is X the same as Z’) across the engines buyers use, on a cadence, logging every stated fact against ground truth.

• Fan-out branch coverage

  Fan-out branch coverage is the classification of brand content against each inferred mid-layer branch in the query fan-out tree — typically scored on a three-state axis: 'yes' (a dedicated sibling page targets the branch as its primary intent with chunk-level rationale alignment on the branch's rationale cluster), 'partial' (a page that mentions the branch but does not target it as the primary intent — partial pages rarely earn cited slots on the branch retrieval), or 'no' (no page targeting the branch at all).

• Fan-out coverage ratio

  The fan-out coverage ratio is the headline metric of a sibling page architecture program: the count of dedicated sibling pages (the 'yes' coverage on the branch scoring) divided by the count of inferred mid-layer branches in the fan-out tree. The ratio is computed per head query and averaged across the priority set.

• Fan-out gap audit

  A fan-out gap audit is the recurring (typically bi-weekly) process that captures the dominant mid-layer branches for the priority head queries, scores brand coverage per branch, computes the coverage ratio, and converts the gaps into the sibling-page editorial backlog.

• Fan-out synthesis stage

  The fan-out synthesis stage is the final composition step where the AI search engine assembles one answer from the per-branch shortlists the retrieval substrate returned across the fan-out tree. Synthesis is not equal-weighted across branches — engines apply a per-branch weighting based on retrieval-rank quality, freshness signal stack, rationale-snippet shape, and multimodal slot availability.

• Favicon trust signal

  Favicon trust signal is the citation-card sub-property that measures how sharply and distinctly the publisher's favicon renders on the AI search citation card. Sharp, distinct favicons rendered at 32×32 or larger source resolution (so the 16×16 and 20×20 downsamples render cleanly) lift card-surface click-through 1.2–1.4× over blurry, generic, or missing favicons.

• Featured snippet

  A featured snippet is a SERP feature that pulls a direct answer — paragraph, list, table, or video clip — from a ranking page and displays it at position zero, above the standard organic results.

• Follow-up retrieval substrate

  The follow-up retrieval substrate is the per-engine pipeline that retrieves candidate chunks for the follow-up turn — biased by the head-turn cited URLs, the head-turn rationale snippets, the head-turn entity slot, and the user's reaction signals from the head-turn answer.

• Fragment integrity

  Fragment integrity is the sentence-level property that measures whether a candidate anchor sentence renders coherently in isolation — as a standalone hyperlink text on the answer surface, without the surrounding chunk.

• Freshness signal stack

  The freshness signal stack is the five-signal set every major AI search substrate reads to compute a page's freshness timestamp: the HTTP Last-Modified header, the schema-emitted dateModified field (Article, FAQPage, Product, HowTo), the visible last-updated date rendered in server-side HTML at the top of the page, the content-diff hash the substrate computes between successive crawls, and (for multimodal pages) the image last-modified plus content hash on the image file.

• Further sources panel

  The further sources panel is the secondary source list AI search engines render below the synthesized answer — chunks that survived retrieval, survived rerank, were read into the synthesis prompt, but were not rendered into the answer's prose with a quoted span or paraphrased sentence.

• GEO (Generative Engine Optimization)

  Generative Engine Optimization (GEO) is the practice of optimizing content so that AI-powered search engines—such as Google AI Overviews, ChatGPT with browsing, Perplexity, and other retrieval-augmented generation systems—cite and surface it in their generated answers.

• Google AI Mode

  Google AI Mode is the conversational, full-page generative-AI experience Google launched in 2025 as a separate tab inside Search. Unlike AI Overviews — which sit at the top of a traditional results page — AI Mode is the entire page: a multi-turn chat interface that synthesizes answers from across the web, supports follow-up questions, and shows source citations as cards beside the response.

• Google-Extended

  Google-Extended is a robots.txt token that lets a site opt out of having its content used to train Google's generative-AI models (Gemini and Vertex AI) — without affecting how Googlebot crawls the site for Google Search or AI Overviews. It is not a separate crawler; it is a control token Googlebot honors.

• GPTBot

  GPTBot is OpenAI's web crawler used to gather content for training and improving its foundation models. It is distinct from OpenAI's answer/retrieval crawlers (OAI-SearchBot and ChatGPT-User), which fetch pages to build cited answers rather than to train — a distinction that matters enormously for AI-search access control.

• Ground-truth surface

  A ground-truth surface is the single canonical page a brand designates as the loud, unambiguous, machine-readable statement of the facts an AI engine keeps getting wrong — current pricing, current features, founding facts, and explicit disambiguation (‘we are not [the other brand]’) where conflation is happening.

• Head-query anchor-word presence

  Head-query anchor-word presence is the sentence-level property that measures whether the priority head query's keyword sits in the first 60% of a candidate anchor sentence. Sentences with the head-query keyword front-loaded in the first 60% win the anchor slot at 1.6–1.9× the rate of sentences that place the keyword in the tail or omit it entirely.

• Heading-bounded passage

  A heading-bounded passage is the block of content between one heading and the next, treated as a single retrievable unit by systems that chunk on structural boundaries. Because many retrieval pipelines split at heading tags, the heading and the text beneath it function together: the heading supplies the topic label the passage is matched against, and the body supplies the answer that gets quoted.

• Hyperlink anchor selection

  Hyperlink anchor selection is the per-engine policy the synthesis stage runs to pick which sentence inside a verbatim-cited chunk becomes the visible hyperlink text on the answer surface. The policy is not published but the rendered anchors converge on stable per-engine patterns through 2026: Google AI Mode anchors on the leading sentence 68% of the time when it carries a numeric or named-entity signal; ChatGPT Search anchors on the highest-numeric-density sentence 58% of the time; Perplexity rotates the anchor across candidate sentences and stabilizes on the highest-engagement sentence over a 4–8 week window; Microsoft Copilot anchors on the freshest-signal sentence 46% of the time; Amazon Rufus runs an asymmetric policy between the product-discovery branch (product-specific sentence) and the use-case branch (scenario-descriptive sentence); Claude anchors on the most reasoning-dense sentence 54% of the time.

• Image freshness window

  The image freshness window is the retrieval-refresh cycle the multimodal-retrieval pipeline runs against for cited carousel slots, materially shorter than the text freshness window. By mid-2026 it runs 4–8 weeks on fast-moving categories (apparel, beauty, food, supplements, accessories), 8–12 weeks on mid-velocity categories (home, fitness, pet, baby), and 12–24 weeks on slow-moving categories (B2B SaaS, financial services).

• Image slot (AI shopping feed)

  An image slot in the AI shopping feed is a discrete, role-tagged image the model consumes for a specific sub-query in the shopper’s intent tree. Mid-2026 AI shopping surfaces converge on four slots: A (hero on white, 1:1 ~2000px, the card thumbnail), B (in-hand at consumer scale, 4:5, the fits signal), C (in-scene use, 4:5 or 9:16, the use-case signal), and D (detail / macro, 1:1, the quality signal).

• ImageObject schema

  ImageObject is the schema.org structured-data type the multimodal-retrieval pipeline reads most heavily for the inline image carousel. The mid-2026 high-leverage properties are contentUrl, caption, name, description, width, height, creator (or author), license, contentLocation, and representativeOfPage. Pages that emit ImageObject inside Product or Article — populated, not stubbed — earn cited image slots in roughly 2.8× the multimodal answers vs. equivalent pages with bare image URLs.

• In-flow checkout

  In-flow checkout is the AI-assistant feature that lets a user complete a purchase inside the assistant’s own surface — no click-through to the brand’s site, no PDP, no third-party cart. OpenAI’s ChatGPT Shopping checkout, Perplexity Shopping’s in-app cart, and Amazon Rufus Buy have all shipped or announced in-flow checkout flows by Q4 2026.

• Inline image carousel

  An inline image carousel is the strip of product or lifestyle images an AI engine surfaces inside the same answer block as its text response — a first-class citation surface alongside the textual citation, not a separate panel or side-rail.

• Knowledge graph

  A knowledge graph is a structured database of entities (people, brands, products, places, concepts) and the relationships between them, used by search engines and AI systems to ground answers in verifiable facts. Google, Microsoft, Apple, and most LLM providers maintain proprietary knowledge graphs that get queried alongside generative retrieval to disambiguate entities and attach authority to claims.

• Knowledge-panel correction

  A knowledge-panel correction is the use of an engine’s or platform’s direct feedback channel — a knowledge-panel edit, a suggest-a-change control, an answer-feedback control, or a structured business-data submission — to correct a wrong fact an engine states about a brand, rather than relying solely on retrieval to override it.

• LLM hallucination

  LLM hallucination is the failure mode where a large language model generates output that sounds confident and well-formed but contains fabricated, incorrect, or unsupported information — invented statistics, made-up product features, miscited sources, nonexistent case law, wrong pricing.

• llms-full.txt

  llms-full.txt is a companion to llms.txt: where llms.txt is a curated Markdown index that points an AI crawler at a site's most important pages, llms-full.txt is the actual concatenated content of those pages as clean Markdown in a single file — the whole 'book' in one download, so a model never has to render the site to read it.

• llms.txt

  llms.txt is a proposed file-based standard (introduced by Jeremy Howard in 2024) that lets a website expose a curated, LLM-friendly summary of its content at the root path /llms.txt. Modeled on robots.txt and sitemap.xml, the file contains markdown-formatted links to the site's most important pages along with one-line descriptions — designed so AI assistants can ingest a clean, structured map of the site without crawling and parsing HTML.

• Local AI search

  Local AI search is the practice of getting a business named in the AI-assistant answers to local-intent queries — ‘best plumber near me’, ‘coffee shop nearby open now’, ‘dentist in [neighborhood] taking new patients’ — as distinct from ranking in the classic map pack.

• Local consensus signal

  The local consensus signal is the locale-specific body of opinion an AI engine reconciles to decide which of the resolvable, constraint-matching local businesses to actually name in a near-me answer — the local analog of the off-site consensus signal that decides national recommendations.

• Local entity graph

  The local entity graph is the set of a business’s location entities — one per physical location — as an AI engine resolves them across map data, business directories, data aggregators, and the open web, before it will consider naming any of them in a local answer.

• Manifest stripping

  Manifest stripping is the loss of a C2PA content credentials manifest during a file-processing step — typically a resize, crop, re-encode, or export path in an editing tool or a platform’s upload pipeline.

• Microsoft Copilot

  Microsoft Copilot is the umbrella brand for Microsoft's AI assistant family — spanning Microsoft 365 Copilot (Word, Excel, PowerPoint, Outlook, Teams), Copilot in Windows, GitHub Copilot, Copilot Studio (custom agents), Bing Copilot (web search and chat), and Copilot in Microsoft Edge.

• Multi-turn citation persistence

  Multi-turn citation persistence is the rate at which a brand retains citation (verbatim, paraphrased, or in the further-sources panel) across consecutive conversation turns within a single AI search session. The metric is the conversation-shaped analog of citation share — share of voice tells you whether you're cited on any single query, while persistence tells you whether you stay cited as the user refines, compares, or pivots within the same conversation.

• Multimodal answer

  A multimodal answer is an AI-engine response that surfaces text, images, and (increasingly) short video clips inline as part of the same answer block — rather than presenting them as separate carousels or side panels.

• Multimodal answer share

  Multimodal answer share is the percentage of priority queries on which a brand holds at least one cited slot inside the AI engine's inline image carousel, weighted by slot position (position 1 = 1.0, position 2 = 0.6, position 3 = 0.3, positions 4–5 = 0.1).

• Multimodal retrieval pipeline

  The multimodal retrieval pipeline is the engine-side image-citation pipeline that fills the inline carousel — distinct from the text-citation pipeline that fills the prose answer. The two run in parallel against partially different signals: text retrieval reads passage density, rationale clarity, entity disambiguation, and source-page freshness, while multimodal retrieval reads ImageObject schema density, image freshness window, alt-text coherence, persona stability across a page set, OG image quality, and product-image accuracy on PDPs.

• NAP consistency

  NAP consistency means rendering a brand's Name, Address, and Phone — and, for digital-first brands, its canonical name, URL, and legal entity — identically everywhere the brand appears online. Originally a local-SEO discipline, it has become a core entity-disambiguation signal for AI search: an engine resolving a brand to a single knowledge-graph node reads inconsistent references as evidence of several weakly-linked entities rather than one strong one.

• Near-me query

  A near-me query is a local-intent question with an implicit or explicit geographic and often constraint component — ‘best X near me’, ‘X in [neighborhood]’, ‘X open now nearby’, ‘emergency X near me’ — that an AI assistant answers by naming specific local businesses rather than returning a list to scroll.

• OAI-SearchBot

  OAI-SearchBot is OpenAI's crawler for its ChatGPT search index — the answer/retrieval crawler that keeps pages available to be surfaced and cited in ChatGPT answers, as opposed to GPTBot, which crawls for model training. A companion user-agent, ChatGPT-User, fetches a page live when a user or agent follows a link inside a chat.

• Offer freshness

  Offer freshness is the recency signal an AI shopping assistant surface reads on a product feed to decide how much weight to give a candidate SKU. It is carried on the feed as an offer_freshness_timestamp on every SKU, refreshed on every feed run — even when nothing has actually changed.

• Open Knowledge Format (OKF)

  The Open Knowledge Format (OKF) is a structured, schema-oriented alternative to the loosely-specified llms.txt convention: a 'knowledge bundle' that packages a brand's key facts, entities, and their relationships as explicit statements a model can ingest without inference — this brand, these products and attributes, these claims with sources, these policies with effective dates.

• Paraphrase citation

  A paraphrase citation is the synthesis-stage outcome where the engine renders the chunk's substance in its own voice — no quoted span — but still attaches a numbered source chip to the paraphrased sentence.

• Passage embedding

  A passage embedding is the engine's vector representation of a single chunk inside a page — not the page as a whole. Every major AI engine through mid-2026 stores one embedding per ~600–900 character chunk inside the retrieval substrate; queries are matched against chunk embeddings, not page embeddings, and the highest-scoring chunk wins the citation.

• Passage ranking

  Passage ranking is a retrieval technique where a search or AI engine ranks individual passages (typically a paragraph or short section) of a long page rather than ranking the page as a whole. Google introduced passage ranking publicly in 2020; by 2026, every major AI search engine relies on passage-level retrieval to decide which span of text to lift into a synthesized answer.

• Passage-level citation

  A passage-level citation is the engine's choice to cite a specific paragraph or sentence on a page rather than the page as a whole. Modern AI engines retrieve and cite at the passage layer — the citation chip resolves to a #:~:text= URL fragment that scrolls the reader to the cited paragraph.

• Passage-level entity grounding

  Passage-level entity grounding is the substrate behavior of resolving named entities (brand names, product names, comparison targets) inside the retrieved chunk against the engine's entity graph — independent of the rest of the page.

• Perplexity AI

  Perplexity AI is an answer-engine search product that combines large language model reasoning with live web retrieval to produce sourced, conversational responses to user queries. Unlike traditional Google-style search, Perplexity returns a synthesized written answer with numbered inline citations rather than a list of blue links.

• PerplexityBot

  PerplexityBot is Perplexity's crawler that maintains the retrieval index feeding its answer engine; a companion user-agent, Perplexity-User, fetches a page live when a user follows a citation. Both feed the answer surface, so a brand that wants to be cited on Perplexity should allow both in robots.txt.

• Persona-locked visual set

  A persona-locked visual set is a page-set's worth of imagery shot or generated with a single recognizable face, body type, styling, and visual identity — the visual analog of a stable named author on the text side.

• Post-retrieval rerank

  Post-retrieval rerank is the cross-encoder scoring pass an AI search engine runs over the retrieved candidate chunk set before passing the top 3–8 chunks to the synthesis stage. The retrieval stage uses fast dense-vector similarity for scale; the post-retrieval rerank stage runs a slower, more expensive cross-encoder on the (sub-query, chunk) pair to read joint semantic and structural properties the embedding stage discounts — claim specificity, named-entity grounding, rationale-shaped opening, freshness stack alignment, schema scaffolding.

• Priority page refresh tier

  A priority page refresh tier is the cadence bucket a page sits in inside an editorial refresh calendar — the tier determines how often the page is refreshed and how the substrate's freshness window is defended on it.

• Product answer eligibility

  Product answer eligibility is the set of gates that determine whether a SKU is allowed to appear in — or be transacted through — an AI answer, independent of how well it matches the query. Discovery eligibility gates whether a product can be cited or surfaced in a carousel; transaction eligibility gates whether an agent can actually buy it.

• Programmatic SEO

  Programmatic SEO (pSEO) is the practice of creating large numbers of SEO-optimized pages by combining a page template with a structured dataset, instead of writing each page by hand. Typical pSEO patterns include: location pages ('plumbers in [city]'), comparison pages ('X vs Y'), alternative pages ('alternatives to X'), category × vertical pages ('UGC for [industry]'), tool calculators (one per niche calculation), and glossary or directory builds.

• Prompt injection

  Prompt injection is a class of security and integrity attack against AI systems where an attacker embeds instructions inside content the AI will read — a webpage, a document, an image, a customer support message — that override the AI's intended behavior when the AI processes that content.

• Provenance chain

  A provenance chain is the ordered sequence of actor-signed actions recorded in a content credentials manifest — for AI UGC, typically c2pa.created (the generator claims the pixels), c2pa.ai_generated (the generator asserts machine origin), c2pa.edited (optional retouching), and c2pa.published (the brand signs off on the final asset).

• Publisher badge recognition

  Publisher badge recognition is the citation-card sub-property that measures whether the AI search engine resolves the publisher entity to a recognizable brand string on the rendered card (rather than collapsing to the bare domain). Publishers that resolve to a recognizable brand string lift card-surface click-through 1.4–1.6× over publishers rendering as bare domains.

• Query fan-out

  Query fan-out is the technique generative search engines (notably Google AI Mode and ChatGPT Search) use to expand a single user query into multiple parallel sub-queries, retrieve sources for each, and synthesize one answer.

• Query fan-out tree

  The query fan-out tree is the three-layer structure an AI search engine's query-expansion stage runs in front of its retrieval substrate: the head query (the string the user typed), the mid-layer sub-queries (the 2–6 parallel expansions the engine generates), and the leaf-layer retrieved chunk sets (per-branch shortlists the substrate returns).

• Rationale pattern cluster

  A rationale pattern cluster is the engine's emergent grouping of citation-rationale snippets by claim type — use-case, comparison, social proof, specification, visual. Pulled across a competitor citation footprint or an own-brand rationale audit, the cluster distribution is the engine's opinion of what counts as a citable answer in your category.

• Rationale snippet

  A rationale snippet is the short justification fragment an AI engine surfaces alongside a citation — the ‘best for sensitive skin’, ‘reviewers report it lasts through a workout’, or ‘works well with the matching toner’ that the engine lifts from the cited source.

• Rationale snippet audit

  A rationale snippet audit is the recurring (typically weekly) process of capturing the rationale snippets AI engines publish alongside their citations, organizing them into a tabular artifact, scoring them on a defined rubric, and converting the audit into a backlog of concrete content rewrites the writers can ship from.

• Reddit citation weighting

  Reddit citation weighting is the disproportionate weight AI engines place on Reddit and forum-style community discussion when sourcing and recommending brands. Two forces compound it: licensing deals that put Reddit content directly into the training and retrieval pipelines of major engines, and years of users appending ‘reddit’ to commercial searches, which taught the models that community discussion is where trustworthy comparison lives.

• Refresh velocity

  Refresh velocity is the count of priority-page refreshes an editorial team ships per week, scored against the refresh cadence the engine freshness window requires across the four-tier calendar. The headline metric is whether refresh velocity is sufficient to defend Tier 1 (every 4–6 weeks) and Tier 2 (every 8–12 weeks) on the program's priority page count — most well-engineered mid-2026 programs ship 6–14 refreshes per week against a 40–120 page priority set.

• Rerank decay pattern

  The rerank decay pattern is the three-phase curve a previously cited chunk rides as rerank survival falls without intervention. Phase 1 — Freshness drift: chunk retrieves but the freshness signal stack has aged into the discount window; rerank survival compresses 25–40% over 8–12 weeks.

• Rerank lift score

  Rerank lift score is the per-chunk projected delta in rerank survival rate from optimizing the chunk against its failing rerank properties — a forward-looking priority score the rerank-edit backlog sorts on.

• Rerank survival rate

  Rerank survival rate is the fraction of brand chunks in the inferred candidate set per priority sub-query that survive the rerank pass and appear in the cited synthesis surface across a rolling 4-week window. It is the headline metric of a rerank-optimization program and the chunk-level analog of citation share.

• Rerank tie-breaker

  A rerank tie-breaker is the secondary signal the cross-encoder applies when two candidate chunks score within ~3% of each other on the primary relevance score — small enough that the primary score alone cannot reliably order them.

• Reranker layer

  The reranker layer is the middle stage in the three-stage retrieval-rerank-synthesis pipeline every major AI search engine runs through 2026. Retrieval returns the top 40–120 candidate chunks per sub-query via embedding similarity; rerank runs a cross-encoder pass that reads each (sub-query, chunk) pair jointly and prunes the candidate set to the top 3–8; synthesis composes the answer from only the reranked top set.

• Retrievable chunk

  A retrievable chunk is a passage that can be selected, understood, and quoted on its own, without the reader or the model needing the surrounding page. Whether a passage qualifies is mostly a writing property rather than a technical one.

• Retrieval substrate

  The retrieval substrate is the underlying corpus, index, and ranking layer an AI engine uses to decide which documents to surface and cite in response to a query. The substrate is engine-specific and not directly inspectable, but its behavior is observable through the citation patterns it produces — which sources it leans on, which content shapes it rewards, which freshness windows it respects.

• Review corpus depth

  Review corpus depth is the volume, recency, and use-case breadth of customer reviews indexable by an AI engine on a brand’s products. AI shopping assistants — Amazon Rufus, Perplexity Shopping, ChatGPT Shopping, Google AI Mode shopping panels — pull rationale snippets directly from the review corpus, so corpus depth has stopped being a vanity metric and become a citation input.

• sameAs property

  The sameAs property is a schema.org attribute that links an entity (most often an Organization or Person) to authoritative external profiles that describe the same entity — Wikidata, Wikipedia, LinkedIn, Crunchbase, official social accounts, and category registries.

• Schema markup

  Schema markup is structured data embedded in a web page — typically as JSON-LD in a <script> tag — that tells search engines and AI assistants what the page is about in machine-readable form. The vocabulary is defined at schema.org, jointly maintained by Google, Microsoft, Yahoo, and Yandex.

• Search intent

  Search intent is the underlying goal a user is pursuing when they type a query — informational ('what is AI UGC'), navigational ('ppl.studio login'), commercial-investigation ('best AI UGC tool 2026'), or transactional ('buy AI UGC subscription').

• Semantic chunk boundary

  A semantic chunk boundary is the point at which a retrieval system splits a document into passages, and where that split lands determines what an engine can retrieve. A boundary that falls cleanly between two complete ideas produces a chunk that stands alone and can be quoted.

• Sentiment corpus

  The sentiment corpus is the body of third-party opinion an AI engine can index about a brand — the reviews, community threads, forum posts, social discussion, and editorial coverage whose aggregate tone and specificity feed the engine’s recommendation decision.

• Share of search (AI)

  Share of search (AI) is the percentage of a category's defined query volume on which a brand is cited, mentioned, or recommended by an AI engine — the AI-search successor to traditional share of search. Unlike share of voice (which counts presence on a curated query set), share of search models the underlying query distribution and weights presence by query volume.

• Share of voice (AI)

  Share of voice (AI) is the percentage of brand-relevant queries inside a defined topic set where a given brand appears in the AI-generated answer — cited as a source, mentioned by name, or recommended by the AI.

• Shortlist position

  Shortlist position is the slot a brand occupies on the 1–5 recommendation list an AI shopping assistant — Amazon Rufus, Perplexity Shopping, ChatGPT Shopping, Google AI Mode shopping panels — surfaces in response to a category-defining query.

• Sibling page architecture

  Sibling page architecture is the editorial-architecture pattern that ships one focused sibling page per dominant mid-layer fan-out branch rather than one comprehensive pillar page for the head query. The pillar still anchors the cluster, but every branch the engine fans into has a dedicated brand-aligned candidate in the per-branch retrieval set — focused intent, chunk-rationale alignment on the branch's dominant rationale cluster, persona-locked visual set on multimodal-active branches, internal links up to the pillar and laterally to peer siblings.

• Signed AI policy

  A signed AI policy is the brand-controlled declaration — published at /robots.txt, /ai.txt, or via the in-development IETF AI Preferences Working Group standard — that names which AI crawlers and assistants are permitted to ingest, index, and cite the site’s content.

• Source freshness window

  The source freshness window is the time horizon inside which an AI engine treats a cited source as preferentially citable rather than discounting it for age. The window is engine-specific, query-class-specific, asymmetric across text and image surfaces, and shortening through 2026.

• Sub-query intent cluster

  A sub-query intent cluster is one of the three dominant intent-slice patterns mid-2026 commercial fan-outs decompose into: specification clusters (the engine isolates a constraint inside the head query — 'under $100', 'for small business', 'without coding' — and runs a sub-query against the constraint alone), use-case clusters (the engine extracts the implied scenario — 'for Shopify', 'for ads', 'for product photos' — and runs a sub-query targeting the scenario), and comparison clusters (the engine runs a named-entity comparison sub-query — 'Tool A vs Tool B', 'Tool A alternatives', 'best Tool A' — to surface competitive candidates).

• Substrate update recovery

  Substrate update recovery is the time it takes a priority page's citation share to return to pre-update levels after the engine ships a substrate update (a re-weighting of the retrieval ranking signals — chunk size, entity grounding, freshness, multimodal slot, rationale-snippet shape).

• Synthesis composition order

  Synthesis composition order is the sequence the synthesis stage uses to assemble the rendered answer from the union of reranked surviving chunks. The engine does not render the answer in citation-numbered order — it composes a fluent answer first, then attaches source chips.

• Synthesis prompt

  The synthesis prompt is the engine-side prompt that takes the union of reranked candidate chunks across every sub-query in the fan-out tree and instructs the LLM to compose a single coherent answer.

• Synthesis stage

  The synthesis stage is the third and final stage in the retrieval-rerank-synthesis pipeline every major AI search engine ran through 2026. The retrieval stage returns 40–120 candidate chunks per sub-query; the reranker stage prunes to 3–8 per sub-query; the synthesis stage takes the union of the surviving chunks across every sub-query in the fan-out tree and composes a single fluent answer.

• Thread-resilient claim shape

  A thread-resilient claim shape is the chunk-level leading-sentence pattern that holds citation across the follow-up turn — a claim that reads as relevant to a category, not just to the head query. The shape extends the citable-claim shape ([Entity] [verb] [quantified claim] [optional qualifier]) by anchoring the claim to the brand-relevant category rather than the specific head-query phrasing.

• Topical authority

  Topical authority is the degree to which a site is perceived — by both search ranking systems and LLM citation systems — as a deep, trustworthy source on a specific subject area, rather than a thin generalist.

• Transcript chunk

  A transcript chunk is a short span of a video's spoken words, with aligned start/end timestamps, treated as an independently retrievable unit by an AI search engine's video pipeline — the video equivalent of a passage in text passage-level optimization.

• Trust block (AI shopping feed)

  The trust block is the segment of an AI shopping feed that carries the ranking-tiebreaker and answer-eligibility signals the model uses to decide which SKUs surface and which get gated.

• Turn-level entity grounding

  Turn-level entity grounding is the explicit naming of the brand or product entity inside the chunk text so the synthesis stage of the follow-up turn re-anchors on the same entity rather than on a generic category placeholder.

• URL template coverage

  URL template coverage is the count of distinct cited URLs a brand has per content template — comparison pages, use-case pages, category pillars, glossary, FAQ, PDPs, case studies, reviews, integration pages, location/service pages. Template coverage is the most actionable axis a citation footprint carries because content teams ship in templates, not in one-offs.

• Vector search

  Vector search is a retrieval technique that finds content by semantic similarity rather than literal keyword match: every document is converted to a high-dimensional numeric vector (an embedding) using an AI model, the query is converted to a vector the same way, and the system returns the documents whose vectors are mathematically closest to the query.

• Verbatim citation

  A verbatim citation is the synthesis-stage outcome where the engine renders a quoted span from the source chunk directly inside the rendered answer, with a numbered source chip attached to the quoted text.

• Video answer slot

  The video answer slot is the inline short clip an AI search engine plays inside an answer to demonstrate what it just explained — the video counterpart to the image carousel in the multimodal answer. The engine does not fill it by ranking whole videos the way classic video SEO did; it retrieves a moment.

• Visual asset matrix

  The visual asset matrix is the (page × image-role) production grid that tracks every image slot a priority page needs filled to compete on the multimodal-answer surface. The mid-2026 standard role list per priority page is three: hero (lead image, product in primary frame), lifestyle (product-in-context shot showing the use case), and detail (close-crop showing texture, finish, or packaging).

• Visual rationale cluster

  A visual rationale cluster is the engine's emergent grouping of carousel-cited images by visual claim type — texture demonstration, product-in-use lifestyle, before-and-after, scale/size reference, packaging/unboxing, ingredient/material close-up. Pulled across a competitor citation footprint or an own-brand multimodal audit, the cluster distribution is the engine's opinion of what counts as a citable image in a category.

• Wikidata

  Wikidata is a free, collaboratively-edited knowledge base of structured facts about entities — people, organizations, products, places — that the major AI search engines lean on heavily for entity attributes and disambiguation.

• Zero-click search

  A zero-click search is a search query that is answered inside the search results page without the user clicking through to a website — the answer is delivered via a featured snippet, knowledge panel, AI Overview, AI Mode response, Map Pack, or similar SERP feature.

Back to the full glossary