The GEO Glossary: 22 Definitions for the Answer-Engine Era

---

A

Answer-Engine. A search counterpart that returns a synthesized answer rather than a ranked list. ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews are the major examples. The defining architecture of the era that began around 2023, when these systems crossed mainstream adoption.

AI Visibility. The degree to which a brand is present, accurate, and on-strategy inside AI-generated answers. The goal of GEO.

C

Citation. Being named, quoted, or referenced inside an AI-generated answer. A new unit of internet authority that operates alongside the hyperlink.

Citation Share. The percentage of AI-generated answers in a defined category that cite a given brand. A standing performance metric for the era. The weighted formula and audit methodology are in What Citation Share Captures (and What It Doesn't).

Credentialed Truth. The source layer AI engines tend to reach for when the cost of being wrong is high. Peer-reviewed research, government agencies, professional associations, named credentialed practitioners. The character of the Expert Layer.

E

Encyclopedia Question. A prompt with a settled, verifiable answer that does not depend on the asker's situation. AI engines tend to ground these prompts on convergent authoritative sources. Contrasts with judgment question. Most real-world prompts are hybrids of the two.

Expert Layer. The credentialed-truth source layer underneath AI answers. Anchored in peer-reviewed publications, government and regulatory bodies, accredited professional associations, and named credentialed practitioners. The layer that tends to carry most weight on high-stakes prompts.

F

Framing Drift. The retrieval penalty AI engines apply to a brand whose public identity changes too often for a stable definition to form, such as a fashion house cycling through creative directors. EPR first documented the term in its coverage of the AI Luxury 25 2026 study, which found Gucci and Balenciaga penalized by LLMs for framing drift tied to creative-director churn and reinvention cycles. See GEO Leaves the PR Trades.

G

GEO (Generative Engine Optimization). The discipline of earning inclusion inside AI-generated answers. Emerged as the answer-engine-era counterpart to SEO.

Grounding. The process by which AI-generated language gets anchored to verifiable, externally validated sources. Different question types tend to activate different grounding behaviors.

Grounding Stack. The five-layer source architecture underneath AI answers: Reddit (judgment), Wikipedia (identity), tier-1 news (recency), expert sources (credentialed truth), and owned properties (self-description). The canonical EPR framework for source-layer architecture.

I

Identity Layer. The structural-definition source layer underneath AI answers, anchored primarily in Wikipedia. The layer AI engines tend to consult first when asked who or what something is.

J

Judgment Layer. The community-consensus source layer underneath AI answers, anchored primarily in Reddit. The layer AI engines tend to lean on for verdicts, comparisons, and lived-experience prompts.

Judgment Question. A prompt with a contingent answer that depends on values, context, and lived experience. Contrasts with encyclopedia question. The class of prompt where messy truth dominates.

M

Messy Truth. The conversational, contested, partially contradictory ground state of human opinion. The source material AI engines tend to prefer for judgment questions because it provides distributional grounding rather than single confident assertions.

N

News Layer. The recency-and-credibility source layer underneath AI answers about the present. Anchored in tier-1 publications: Reuters, Bloomberg, the Financial Times, the Wall Street Journal, the Associated Press, the New York Times, and their equivalents. Coverage tends to depreciate, with the past quarter weighted heavier than older coverage on time-sensitive prompts.

Notability Threshold. Wikipedia's standard for entry creation: significant coverage in reliable independent sources. The entry point to the Identity Layer. Cannot be manufactured.

O

Owned Layer. The self-description source layer underneath AI answers, the brand's own documentation, pricing pages, help content, primary-source identity. The only layer the brand fully controls. The layer that tends to be weighted least for evaluative claims.

R

Retrieval Anchor. A piece of content AI engines reach for when grounding an answer. The unit of source material inside GEO.

S

Source Ecosystem. The full connected universe of sources AI engines retrieve from when generating answers, spanning the five source layers of the Grounding Stack. The thing brands build presence inside through GEO. Sometimes referred to as the underlying source base or retrieval substrate.

T

Tier-1 Publications. The small set of news outlets AI engines tend to weight disproportionately inside the News Layer: Reuters, Bloomberg, the Financial Times, the Wall Street Journal, the Associated Press, the New York Times, and a handful of comparable peers.

Trust Discount. The structural devaluation AI engines tend to apply to evaluative claims a brand makes about itself. Owned-content claims that the brand is "leading" or "best" tend to carry less weight than the same claim made by a third party.

Related EPR Coverage

• The GEO Canon
• GEO in One Page: A Founder's Definition
• The Citation Share Index
• The GEO Operating Stack: 14 Layers
• The AI Communications 100
• What Citation Share Captures
• The 35-Prompt Citation Share Audit
• GEO Leaves the PR Trades: Tech Press Picks Up the Story

Frequently Asked Questions

What is the GEO Glossary?

Everything-PR's working dictionary of Generative Engine Optimization terminology: 22 definitions covering the answer-engine, citation mechanics, source-layer architecture, and the disciplines underneath how ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews answer questions.

What is the Grounding Stack?

The five-layer source architecture underneath AI answers: Reddit (judgment), Wikipedia (identity), tier-1 news (recency), expert sources (credentialed truth), and owned properties (self-description). The canonical EPR framework for source-layer architecture.

What is the difference between an encyclopedia question and a judgment question?

An encyclopedia question has a settled, verifiable answer, such as the capital of France or a company's founding date. AI engines ground these on convergent authoritative sources. A judgment question has a contingent answer that depends on values, context, and lived experience, such as which vendor is best or whether a brand is worth it. AI engines lean on distributional sources like Reddit. Most real-world prompts are hybrids.

What is a retrieval anchor?

A piece of content AI engines reach for when grounding an answer. The unit of source material inside GEO. Retrieval anchors can be Wikipedia entries, tier-1 news articles, Reddit threads, regulatory filings, or original research, whichever the engine treats as authoritative for that prompt class.

What is framing drift?

The retrieval penalty AI engines apply to a brand whose public identity changes too often for a stable definition to form. EPR documented the pattern in luxury fashion, where creative-director churn at houses including Gucci and Balenciaga produced measurably lower AI visibility scores in the AI Luxury 25 2026 study.

Why are owned-content claims worth less than third-party claims?

Because of the Trust Discount. AI engines apply a structural devaluation to evaluative claims a brand makes about itself. A brand calling itself "leading" or "best" on its own pricing page carries less weight than the same claim made by Reuters, Forbes, or a credentialed analyst. This is the single most important reason a large majority of AI citations come from earned media rather than brand blogs.