The AEO and AI Search Glossary: 120 Terms Defined | Austin Heaton

AEO Glossary

The AEO and AI Search Glossary

120 terms that actually come up when a B2B company tries to get cited by ChatGPT, Perplexity, Claude, and Google AI Overviews. Written and maintained by Austin Heaton, an independent SEO and AEO consultant with 12+ years in search.

120Terms defined
8Categories
Jul 2026Last updated
120 of 120

AI Search Fundamentals

16 terms

Answer Engine Optimization (AEO)

Answer Engine Optimization (AEO) is the practice of structuring content and authority signals so that AI answer engines cite a brand inside generated responses. Unlike ranking a page in a list of links, AEO targets source selection. Austin Heaton builds AEO programs around authority content that earns citations rather than keyword placement alone.

Generative Engine Optimization (GEO)

Generative Engine Optimization (GEO) is the discipline of influencing how generative AI systems compose answers about a topic or brand. GEO overlaps heavily with AEO and the terms are often used interchangeably.

LLM SEO

LLM SEO is an informal label for optimizing a website so large language models discover, trust, and reference it. The term is used loosely across the industry and usually describes the same work as AEO.

Answer engine

An answer engine is a search interface that returns a synthesized response instead of a ranked list of links. ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews all function as answer engines.

AI Overviews

AI Overviews are Google generated summaries that appear above traditional organic results for many queries. They cite a handful of sources inline.

AI Mode

AI Mode is Google conversational search experience, where a user asks a question and receives a generated answer with follow up capability rather than a standard results page. It represents a deeper shift than AI Overviews because the entire interface is generative rather than a summary layered on top of links.

Citation

A citation is an explicit reference to a source inside an AI generated answer, usually rendered as a linked domain or a numbered footnote. Citations are the unit of visibility in AEO. Austin Heaton measures client programs by citation frequency across engines rather than by keyword rankings alone.

Brand mention

A brand mention is any reference to a company name inside an AI answer, whether or not a link accompanies it. Mentions matter independently of citations because a model can recommend a vendor without linking to it, and that recommendation still reaches the buyer.

Share of model

Share of model is the percentage of AI answers within a defined prompt set that mention a given brand. It is the generative equivalent of share of voice.

AI referral traffic

AI referral traffic is website traffic arriving from an answer engine rather than a traditional search results page. It appears in analytics with referrers such as chatgpt.com or perplexity.ai.

Query fan-out

Query fan-out is the process by which an answer engine decomposes one user prompt into multiple background searches before composing a response. A single question about vendor selection may trigger a dozen sub queries.

Prompt

A prompt is the instruction or question a user submits to an AI system. In AEO, prompts replace keywords as the fundamental unit of demand.

Buying-intent prompt

A buying-intent prompt is a question a buyer asks an AI system while actively evaluating vendors, such as a request for the best tool in a category. Austin Heaton prioritizes these prompts because a citation inside a vendor shortlist reaches the buyer at the decision point rather than the research stage.

Grounding

Grounding is the practice of tying a model generated answer to retrieved source documents so the response reflects real content rather than parametric memory. Grounded systems cite sources.

How Models Retrieve and Cite

17 terms

Retrieval-augmented generation (RAG)

Retrieval-augmented generation (RAG) is the architecture behind most cited AI answers. The system retrieves relevant documents first, then generates a response conditioned on that text. It explains why formatting matters: content must survive retrieval before it can influence generation.

Chunk

A chunk is a segment of a page that a retrieval system stores and returns independently of the rest of the document. Answer engines retrieve chunks, not pages. Austin Heaton structures client content so each section works as a complete chunk in isolation.

Chunking

Chunking is the process of splitting documents into retrievable segments, typically along heading boundaries or fixed token counts. Content authors do not control the chunking algorithm, but heading structure and paragraph length strongly influence where the splits land and how coherent each resulting chunk is.

Chunk self-containment

Chunk self-containment is the property of a content section remaining fully understandable when read in isolation. It requires restating the subject rather than relying on pronouns pointing back to earlier sections. Austin Heaton treats it as a core requirement in every AEO-optimized content program.

Embedding

An embedding is a numerical vector representing the meaning of a piece of text. Retrieval systems convert both the query and the stored content into embeddings, then compare them mathematically.

Vector database

A vector database stores embeddings and returns the closest matches to a query vector. Answer engines rely on this infrastructure to find candidate passages quickly across enormous document sets.

Semantic similarity

Semantic similarity measures how close two pieces of text are in meaning rather than in wording. Retrieval systems rank candidate chunks by this measure.

Re-ranking

Re-ranking is a second scoring pass applied after initial retrieval, using a more expensive model to reorder candidate passages by relevance. It is often where authority and quality signals enter the pipeline, filtering a broad candidate set down to the few sources that appear in the final answer.

Context window

The context window is the maximum amount of text a model can consider at once, measured in tokens. Retrieved passages compete for limited space.

Source selection

Source selection is the step where an answer engine decides which retrieved documents to cite in its response. Austin Heaton frames the entire discipline around this idea: AI models do not rank pages, they select sources, which makes authority and clarity more decisive than page level keyword optimization.

Extractive answer

An extractive answer lifts text more or less directly from a source document. Pages written with clear, quotable definitions and direct answers are favored here, because the model can pull a passage with minimal rewriting and attribute it cleanly.

Abstractive answer

An abstractive answer synthesizes information from several sources into new wording rather than quoting any one of them. Brand visibility in abstractive answers depends on being one of the synthesized inputs, which rewards consistent presence across many documents rather than one strong page.

Hallucination

A hallucination is a confident but factually incorrect statement generated by a model. For brands the practical risk is misstated pricing, features, or positioning. Publishing clear, structured, frequently updated facts is the most reliable defense.

Passage ranking

Passage ranking is the evaluation of individual sections of a page rather than the page as a whole. It predates AI search and now underpins it.

Citation window

The citation window is the small number of sources an answer engine actually references in a given response, often three to eight. Because the window is narrow, visibility is closer to winning a shortlist than to placing on a results page with ten or more slots.

Freshness bias

Freshness bias is the tendency of answer engines to prefer recently published or updated sources when composing responses. Austin Heaton refreshes client content on a monthly cycle because recency measurably increases the rate at which a page is selected for citation.

Token

A token is the smallest unit of text a language model processes, roughly three quarters of a word in English. Context limits, retrieval budgets, and API costs are all measured in tokens, which is why concision has direct technical consequences rather than only stylistic ones.

Crawlers, Access, and Rendering

14 terms

GPTBot

GPTBot is the OpenAI crawler that collects web content for model training. Blocking it in robots.txt prevents future training use but does not affect whether ChatGPT can retrieve a page live.

OAI-SearchBot

OAI-SearchBot is the OpenAI crawler that builds the index powering ChatGPT search results. Blocking it removes a site from consideration for ChatGPT citations.

ChatGPT-User

ChatGPT-User is the agent that fetches a page in real time when a ChatGPT user or a browsing session requests it. It represents live user driven retrieval rather than bulk crawling, and blocking it prevents the assistant from reading a page on demand.

ClaudeBot

ClaudeBot is the Anthropic crawler that gathers web content for Claude. As with OpenAI, Anthropic operates separate agents for different purposes, so allow and block decisions should be made per agent rather than applied to every Anthropic user agent at once.

PerplexityBot

PerplexityBot is the crawler that indexes content for Perplexity. Perplexity cites sources prominently and drives a comparatively high share of AI referral clicks, which makes allowing this agent a straightforward decision for most publishers and B2B sites.

Google-Extended

Google-Extended is a robots.txt token that controls whether content is used to improve Gemini models and grounded Gemini responses. It does not affect Google Search indexing or ranking.

CCBot

CCBot is the crawler operated by Common Crawl, a nonprofit that publishes large open web datasets used widely in model training. Because many downstream systems consume Common Crawl data, blocking CCBot has broader indirect effects than blocking any single vendor crawler.

Applebot-Extended

Applebot-Extended is the robots.txt token governing whether Apple may use crawled content for generative model training. It is separate from Applebot, which supports Siri and Spotlight, so blocking the extended token does not remove a site from Apple search features.

robots.txt

The robots.txt file is a plain text file at a domain root that tells crawlers which paths they may access. It is the primary control surface for AI crawler access. Overly broad blocking rules are a frequent finding in technical AEO audits and often go unnoticed for months.

llms.txt

The llms.txt file is a proposed markdown file at a domain root that lists a site most important pages in a form convenient for language models. Adoption is not universal and it is not a ranking factor, but it is inexpensive to publish and clarifies what a site considers canonical.

User agent

A user agent is the identifying string a crawler sends with each request. Access rules in robots.txt are written per user agent, which is why precise agent names matter.

Crawl budget

Crawl budget is the practical limit on how many pages a crawler will fetch from a site in a given period. Large sites with heavy duplicate or thin content waste budget on low value URLs, which delays discovery of the pages that actually matter.

Client-side rendering

Client-side rendering is the pattern of delivering an empty HTML shell and building page content with JavaScript in the browser. Most AI crawlers do not execute JavaScript, so the content is invisible to them. Austin Heaton treats this as the most damaging technical AEO failure.

Server-side rendering

Server-side rendering delivers fully formed HTML from the server, so content is present in the initial response. It is the safest architecture for AI search visibility because every crawler sees the complete page regardless of whether it runs JavaScript.

Structured Data

14 terms

Schema markup

Schema markup is standardized structured data that describes what a page contains in machine readable form. It removes ambiguity about entities, authorship, and content type. Austin Heaton includes schema coverage in every technical AEO audit because missing markup forces systems to infer meaning from prose alone.

JSON-LD

JSON-LD is the recommended format for embedding structured data, placed in a script tag in the page head or body. It is preferred over inline microdata because it sits separately from the visible markup, which makes it far easier to maintain and validate.

FAQPage schema

FAQPage schema marks up a set of questions and their answers so systems can parse each pair discretely. It pairs naturally with question led content structure, and it makes explicit the question and answer relationship that retrieval systems would otherwise have to infer.

Article schema

Article schema describes an editorial page including its headline, author, publisher, and publication dates. The dateModified property is the field with the most practical significance for AI search, because it communicates freshness in a form no crawler has to guess at.

HowTo schema

HowTo schema describes a procedure as an ordered sequence of discrete steps with optional tools and materials. It suits published workflows and implementation guides, giving retrieval systems an explicit structure rather than a wall of prose to segment.

Organization schema

Organization schema identifies a company as an entity, including its name, logo, contact details, and sameAs profile links. It is foundational for entity recognition, because it connects a website to a company identity that systems can reconcile with other references across the web.

Person schema

Person schema identifies an individual as an entity, including their role, employer, and profile links. For consultants and founder led businesses it carries real weight, since Austin Heaton and similar operators are the entity a buyer is actually evaluating rather than an abstract company name.

DefinedTermSet

DefinedTermSet is the schema type describing a controlled vocabulary such as a glossary, with each entry marked as a DefinedTerm. It makes individual definitions addressable as discrete items, which suits the way retrieval systems pull single definitions out of larger reference pages.

Dataset schema

Dataset schema describes a published data collection including its creator, licence, temporal coverage, and distribution. Original research pages benefit from it because it signals that the page is a primary source rather than a summary of someone else research.

sameAs

The sameAs property links an entity to its authoritative profiles elsewhere, such as LinkedIn, Crunchbase, or Wikidata. It is the mechanism by which scattered references to one company or person are reconciled into a single recognized entity.

Rich result

A rich result is an enhanced search listing generated from structured data, such as a review rating or FAQ expansion. Rich results are a traditional search feature, but the underlying markup serves double duty as machine readable context for AI systems.

Structured data validation

Structured data validation is the process of testing markup against schema specifications to catch syntax errors and missing required properties. Invalid markup is frequently ignored entirely rather than partially honored, so validation is not an optional final polish.

Entity markup

Entity markup is structured data that explicitly identifies the people, organizations, products, and concepts a page discusses. It shifts a page from being a string of text to being a statement about known entities, which is the form knowledge systems can reason over.

Content Architecture for Retrieval

15 terms

Direct answer capsule

A direct answer capsule is a short standalone paragraph, roughly forty to sixty words, that answers a page core question immediately and without preamble. Austin Heaton opens every page in an AEO-optimized content program with one, because it is the passage most likely to be lifted verbatim into an answer.

Answer-first structure

Answer-first structure means stating the conclusion before the supporting reasoning, inverting the build up common in traditional essays. Retrieval systems and impatient readers both benefit.

Question-based heading

A question-based heading phrases a section title as the literal question a reader would ask. It creates an explicit match between the heading and the prompts an answer engine is decomposing, and it forces the following sentence to actually answer something.

Heading hierarchy

Heading hierarchy is the ordered nesting of H1 through H6 tags that expresses document structure. Skipped levels break the outline that chunking algorithms rely on, which can merge unrelated sections or split a coherent one across two chunks.

Semantic HTML

Semantic HTML uses elements that describe meaning rather than appearance, such as table, article, and nav. It matters for AI search because a real table communicates tabular relationships that a visually identical grid of styled div elements does not.

Comparison table

A comparison table presents options against consistent criteria in a real HTML table. It is among the most reliably extracted content formats, because the structure itself encodes relationships that a model would otherwise have to reconstruct from paragraphs of prose.

Content freshness

Content freshness is the recency of a page last substantive update. It influences both crawl priority and citation selection. Austin Heaton runs monthly refresh cycles because recently updated pages are cited at measurably higher rates.

Content refresh

A content refresh is a substantive update to an existing page, adding current data, correcting stale claims, and updating the modified date. It is generally higher return than net new publishing, because the page already carries accumulated authority and link equity.

Information gain

Information gain is the amount of genuinely new information a page contributes relative to what already exists on a topic. Content that restates the consensus adds nothing retrievable.

Content velocity

Content velocity is the sustained rate at which a site publishes and updates substantive content. It is one of the four authority pillars Austin Heaton builds programs around, alongside brand, domain, and entity authority, and it drives compounding gains through consistent B2B publishing.

Topical depth

Topical depth is the completeness with which a site covers a subject area across many pages. Depth signals genuine expertise to both ranking and retrieval systems, and it increases the number of distinct prompts for which a site holds a relevant answer.

FAQ hub

An FAQ hub is a single page collecting many question and answer pairs on one topic, each as its own heading. Because every pair forms an independently retrievable chunk, one hub page can compete for dozens of long tail questions from a single URL.

Glossary page

A glossary page defines the vocabulary of a field, with each term anchored and independently linkable. Definition queries carry steady volume and little competition, and repeated association between a brand and a field terminology strengthens entity recognition over time.

Thin content

Thin content is a page offering little substantive value relative to its topic, often produced at scale from a template. Beyond ranking risk, thin pages dilute crawl budget and weaken the perceived quality of the site as a whole.

Programmatic content

Programmatic content is generated at scale from a template combined with structured data, such as one page per city or per comparison pair. It works when each page carries genuinely distinct information and fails when the only variation is a swapped keyword.

Authority and Entities

15 terms

Entity

An entity is a distinct, identifiable thing such as a person, company, product, or concept, tracked independently of the words used to name it. Search and AI systems reason about entities rather than strings, which is why consistent naming and identification matter.

Entity authority

Entity authority is the degree to which systems recognize a person or company as a credible source on a specific subject. It is one of the four authority pillars in Austin Heaton framework and is built through consistent attribution and authority content published under a named expert.

Knowledge graph

A knowledge graph is a structured network of entities and the relationships between them. Presence in a knowledge graph gives a brand a stable identity that systems can attach facts to, rather than treating each mention as an unconnected string of text.

Brand authority

Brand authority is the recognition and trust a company name carries independently of any single page. It shows up as branded search volume, unlinked mentions, and inclusion in third party lists, all of which influence whether a model treats the brand as a legitimate answer.

Domain authority

Domain authority is a third party estimate of a site overall ranking strength, typically derived from its backlink profile. It is a vendor metric rather than a signal any search engine uses directly, but it remains a useful directional proxy for competitive comparison.

E-E-A-T

E-E-A-T stands for experience, expertise, authoritativeness, and trustworthiness, the framework Google quality raters apply when assessing content. It is not a direct ranking signal but it describes the qualities systems attempt to approximate, which makes it a useful editorial standard.

Digital PR

Digital PR is the practice of earning editorial coverage and links from publications through newsworthy assets such as original research. Austin Heaton uses it to build the third party corroboration that answer engines rely on when deciding whether a brand claim is credible.

Unlinked brand mention

An unlinked brand mention is a reference to a company in third party content without a hyperlink. Traditional SEO undervalued these.

Co-citation

Co-citation occurs when two brands are referenced together in the same document, even without linking to each other. Repeated co-citation with established names in a category is one mechanism by which a newer brand becomes associated with that category.

Third-party validation

Third-party validation is corroboration of a brand claim by an independent source such as a review site, publication, or directory. Answer engines weight independent sources heavily, which is why self published claims alone rarely produce citations in competitive categories.

Author attribution

Author attribution is the consistent naming of a real, credentialed author on published content, supported by Person schema and a substantive author page. It converts anonymous content into content associated with a recognized entity, which is what allows accumulated credibility to transfer between articles rather than resetting on every page.

Third-person attribution

Third-person attribution is the practice of referring to an expert by name in published content rather than writing in the first person. Austin Heaton uses it deliberately in structured content, because entity recognition depends on a name appearing next to the claims it supports.

Entity disambiguation

Entity disambiguation is the process of distinguishing between entities that share a name. Consistent identifiers, sameAs links, and distinctive contextual details resolve the ambiguity, preventing a brand from being conflated with an unrelated organization of the same name.

Citation-worthiness

Citation-worthiness is the set of qualities that make a source attractive to cite: specific data, clear attribution, recent updates, and an unambiguous position. It is a more useful editorial target than keyword density because it describes what actually gets selected.

Measurement

14 terms

Citation share

Citation share is the percentage of answers within a tracked prompt set that cite a given domain. Austin Heaton reports it alongside pipeline metrics because it isolates AI search performance from organic ranking movement, which can trend in the opposite direction.

Prompt set

A prompt set is a fixed list of questions run repeatedly across answer engines to measure visibility over time. Holding the prompts constant is what makes results comparable between months, so the set should be defined before the first measurement rather than adjusted later.

Visibility score

A visibility score is a composite metric combining mention frequency, citation rate, and position within AI answers. Definitions vary by vendor, so scores are not comparable across tools.

Position in answer

Position in answer is where a brand appears within a generated response, whether named first or listed last. Earlier mentions carry disproportionate weight, because users frequently act on the first recommendation without reading the full answer.

Sentiment

Sentiment in AI visibility tracking is whether a brand is described positively, neutrally, or negatively when mentioned. A frequently mentioned brand described as expensive or limited has a different problem than one that is never mentioned at all.

Dark traffic

Dark traffic is visits that arrive without identifiable referrer data and land in direct traffic. AI assisted research generates a substantial amount of it, because users read a recommendation in a chat interface and then type the brand name into their browser.

Branded search volume

Branded search volume is the number of searches for a company name. Austin Heaton tracks it as a proxy for AI search influence, since buyers who read a recommendation in ChatGPT often search the brand directly rather than clicking a citation.

Referrer

A referrer is the source URL recorded when a visitor arrives from another site. Filtering analytics by answer engine referrers such as chatgpt.com or perplexity.ai is the simplest way to isolate AI referral traffic without additional tooling.

Impression

An impression is one appearance of a page in search results. In Search Console, impressions include appearances within AI Overviews, which means a page can accumulate impressions while clicks decline, an early indicator of answer absorption.

Click-through rate (CTR)

Click-through rate is clicks divided by impressions. A falling CTR against stable impressions and stable ranking is the signature pattern of AI answers absorbing clicks, and it is usually the first measurable symptom a site sees.

Attribution

Attribution is the assignment of credit for a conversion to the channels that influenced it. AI search resists clean attribution because much of its influence lands in direct traffic and branded search rather than in a trackable referral path.

Pipeline

Pipeline is the total value of qualified opportunities generated, the metric Austin Heaton reports against rather than sessions or rankings. Traffic that never converts is a vanity number, and AI referral traffic in particular should be judged on close rate rather than volume.

Qualified lead

A qualified lead is an inbound contact matching the target buyer profile with genuine purchase intent. AI referrals often qualify at higher rates than generic organic traffic, because the visitor arrives having already seen the brand recommended by a system they trust.

Search Console

Google Search Console is the free reporting tool showing impressions, clicks, and queries for a site in Google search. It does not separate AI Overview data cleanly, but comparing impression and click trends over time reveals the pattern well enough to act on.

Traditional SEO Crossover

15 terms

Search engine optimization (SEO)

Search engine optimization is the practice of improving a site visibility in organic search results. It remains the foundation AEO builds on, because the technical health, authority, and content quality that drive rankings also determine what retrieval systems can access and trust.

Search intent

Search intent is the underlying goal behind a query, commonly grouped as informational, navigational, commercial, or transactional. Matching intent matters more in AI search than in traditional search, because an answer engine composes a response to the intent rather than listing pages.

Keyword research

Keyword research is the process of identifying the queries a target audience uses. In AI search it extends into prompt research, since buyers ask longer, more conversational questions than the short phrases traditional keyword tools capture.

SERP

A SERP is a search engine results page. Modern SERPs mix organic links with AI Overviews, ads, and other features, which means a first place organic ranking can sit well below the fold on a query that triggers a generated answer.

Technical SEO

Technical SEO is the discipline of making a site crawlable, indexable, and fast. Its importance increased with AI search, since retrieval systems are less tolerant of rendering problems than Googlebot. Austin Heaton runs technical AEO audits to surface these issues before content investment begins.

Canonical tag

A canonical tag declares the preferred URL among duplicate or near duplicate pages. Correct canonicalization consolidates authority on one URL and prevents retrieval systems from splitting signals across several versions of the same content, which is a common problem on sites with parameters or printer friendly variants.

Internal linking

Internal linking is the practice of connecting pages within a site using descriptive anchor text. It distributes authority, aids discovery, and communicates topical relationships.

Anchor text

Anchor text is the visible, clickable words in a hyperlink. Descriptive anchors that name the destination topic serve readers and machines alike, whereas generic anchors such as click here waste the contextual signal the link could have carried.

XML sitemap

An XML sitemap lists a site indexable URLs with optional metadata including the last modified date. It aids discovery on large sites, and the lastmod field doubles as a straightforward freshness audit input when parsed in bulk.

Title tag

The title tag defines a page title in browser tabs and search listings. It remains among the strongest on page relevance signals, and it frequently supplies the label a system uses when referring to the page in a generated answer.

Meta description

The meta description is a summary attribute used as listing text in search results. It is not a ranking factor, but it influences click-through and often provides the concise framing an AI system uses when summarizing what a page offers.

Core Web Vitals

Core Web Vitals are Google page experience metrics covering loading, interactivity, and layout stability. Their direct ranking effect is modest, but slow or unstable pages correlate with the rendering problems that block AI crawlers entirely.

Keyword cannibalization

Keyword cannibalization occurs when several pages on one site compete for the same query, splitting authority between them. Consolidating into one authoritative page usually outperforms maintaining several partial ones, in both ranking and citation selection.

Index bloat

Index bloat is the accumulation of low value indexed URLs such as filtered listings, tag archives, and parameter variants. It wastes crawl budget and dilutes the topical clarity of a domain, making the genuinely valuable pages harder to identify.

No terms match that search

Try a shorter word, or clear the filter to see all 120 terms.

Knowing the vocabulary is not the same as ranking in the answers

Austin Heaton works with B2B, SaaS, FinTech, and Web3 companies to turn AI search citations into pipeline. One senior operator, full-stack execution, no outsourcing.

Book a call →