Every term in the AI-search ecosystem in one place, defined in plain English, with cross-references to the deeper explainer pages where it makes sense.
This page is organized alphabetically. If you're new to GEO, start with What is GEO? first, then come here for the definitions.
A
AEO (Answer Engine Optimization) A term used interchangeably with GEO by some practitioners. Refers to optimizing for AI engines that produce direct answers (not just lists of links). GEO is the more widely-adopted term in 2026.
AI Mode Google's conversational AI search experience launched in 2025, separate from AI Overviews. Powered by Gemini. Has its own retrieval logic and citation patterns distinct from traditional Google Search.
AI Overviews The AI-generated summary that appears at the top of Google Search results for many queries. Powered by Gemini. Approximately 50% of Google searches showed AI Overviews by mid-2025, projected to rise to 75%+ by 2028 per McKinsey.
AI Visibility The umbrella term for how often, where, and how prominently your brand appears across AI engines. Measured through three sub-metrics: citation rate, mention rate, and share of voice. See Citations vs mentions vs links.
Anthropic The company behind Claude. Founded in 2021, focused on AI safety research. Headquartered in San Francisco.
Attribution (in AI search) The process of matching AI citation or mention events to actual customer actions (orders, signups, conversions). Notably harder than traditional click-attribution because the buying decision often happens inside the AI answer, before any click.
B
Bing Index Microsoft's search index. The retrieval backbone for Copilot and a meaningful contributor to ChatGPT's web-mode searches (roughly 87% of ChatGPT's search-mode citations match Bing's top 10 organic results).
Brand mention When an AI engine names your brand in a response without providing a clickable link. Roughly 70% of brand appearances in AI answers are mentions rather than citations.
C
Citation (in AI search) When an AI engine names your brand and provides a clickable link to your domain in the response. Citations are roughly 15% of brand appearances on average, but vary by engine — Perplexity cites more aggressively than ChatGPT.
Citation rate The percentage of AI responses to a defined set of relevant prompts that include a clickable URL to your domain. One of three core AI visibility metrics, alongside mention rate and source rate.
Claude Anthropic's family of large language models. Strong in long-context reasoning, structured analysis, and research-heavy queries. Reaches commerce intent primarily through professional and considered-purchase research.
Co-citation When your brand is mentioned alongside competitors in the same AI response. Co-citation patterns reveal how AI engines view your competitive set — sometimes very differently from how you view it yourself.
Compliance language Wording that meets regulatory standards for a given category (FDA for supplements, FTC for advertising, etc.). AI engines actively deprioritize content using flagged non-compliant claim language, especially in YMYL categories.
Copilot (Microsoft) Microsoft's conversational AI, deeply integrated into Windows, Microsoft 365, and Bing. Uses the Bing index as its primary retrieval source.
D
Decay (content decay in AI search) The phenomenon of older content losing citation priority over time, even when content quality hasn't changed. AI engines weight freshness, so unmaintained content gets progressively deprioritized. Typically begins 60–90 days after publication without updates.
Domain authority (in AI context) A loose term for how much an AI engine trusts a given domain as a source. Unlike traditional SEO's domain authority metric, this is not directly measurable — but high-citation-rate domains tend to share traits: original content, fact density, third-party corroboration, and consistent freshness.
E
Embedding A vector representation of text used by LLMs to find semantically related content. Behind the scenes, when an AI engine searches its retrieval index, it's matching embeddings rather than literal keywords. This is why semantically rich, well-defined content outperforms keyword-stuffed content for GEO.
EWG Verified A third-party certification from the Environmental Working Group with formal verification criteria, distinct from the looser "EWG-friendly" label. Carries meaningful weight in clean-beauty AI citations.
F
FAQPage schema Schema.org markup that identifies a section as a Frequently Asked Questions block. AI engines extract questions and answers from FAQPage-marked content with high reliability. One of the highest-leverage structured data interventions for GEO.
Fact density The ratio of specific, verifiable information (numbers, dates, names, percentages, dimensions) to total content in a passage. AI engines preferentially cite high-fact-density content. Content cited by Perplexity, in one 2024 SEJ analysis, contained 32% more explicit definitions than uncited content.
Freshness A signal AI engines use to evaluate whether content is current. New content typically enters citation pools within 3–5 business days. Updated content with visible timestamps gets re-prioritized. Stale content decays.
G
Gemini Google's family of large language models. Powers AI Overviews, AI Mode, and Workspace AI features. Heavily grounded in Google Search and the Knowledge Graph. Most authority-leaning of the major AI engines.
GEO (Generative Engine Optimization) The practice of structuring content, products, and digital presence so AI engines cite and recommend you in their answers. See What is GEO?.
Grounding The process of constraining an AI model's response to information retrieved from authoritative sources, rather than allowing it to generate freely from training data. Gemini "grounds with Google Search"; Perplexity grounds with its real-time web retrieval.
H
Hydrogen Shopify's headless commerce framework, based on Remix. Allows full control over storefront rendering, which is useful for serving custom structured data, llms.txt, and AI-optimized PDP templates.
I
INCI (International Nomenclature of Cosmetic Ingredients) The standardized naming system for cosmetic ingredients. AI engines extract INCI lists for beauty queries; publishing full INCI for every product is foundational for beauty GEO.
Inference The process of an LLM generating a response. Different from training. When an AI engine answers a query, it's running inference on its retrieved sources.
K
Knowledge Graph Google's structured database of entities and their relationships. Gemini uses Knowledge Graph entries to ground responses. Having an accurate Knowledge Graph entry for your brand meaningfully improves Gemini citation likelihood.
L
Large Language Model (LLM) The type of AI model that powers ChatGPT, Claude, Gemini, Copilot, Perplexity, and others. Trained on vast amounts of text; generates responses by predicting likely sequences of tokens. The "engines" in "Generative Engine Optimization."
llms.txt A Markdown file published at the root of a domain that summarizes a site's content for LLMs. Proposed by Jeremy Howard in late 2024. See llms.txt for Shopify stores.
llms-full.txt A longer companion file to llms.txt containing the full text of a site's highest-priority content concatenated. Optional but useful for sites with substantial original research or long-form content.
M
Mention rate The percentage of AI responses to a defined set of prompts that name your brand, with or without a clickable link. Generally the most important single AI visibility metric for revenue impact.
Metafield (Shopify) A typed, structured data field attached to a Shopify product, collection, or other resource. The primary surface for exposing product attributes to AI engines via Schema.org markup. See Shopify metafields for AI parsing.
Metaobject (Shopify) A newer Shopify construct that lets you define custom data structures (ingredient libraries, certification libraries, etc.) that can be referenced from multiple products. Useful for managing controlled vocabularies at scale.
O
OpenAI The company behind ChatGPT. Founded in 2015, headquartered in San Francisco. ChatGPT has approximately 700M+ weekly active users as of early 2026.
P
Perplexity A RAG-first AI engine focused on transparent, citation-heavy responses. Cites 2–3× more sources per response than ChatGPT. Roughly 30M weekly active users; highest commerce intent per query of the major engines.
Prompt The text input a user gives to an AI engine. In GEO, "prompts" usually refer to the queries that drive commerce intent in your category — the things shoppers ask AI when researching products.
Prompt intelligence The practice of mapping every high-intent prompt in a category and matching them to specific products. The first step in a serious GEO program.
R
RAG (Retrieval-Augmented Generation) The technique most modern AI engines use to answer queries with current information: first retrieve relevant sources from a search index, then generate a response grounded in those sources. Perplexity is the most RAG-pure of the major engines; ChatGPT, Claude, and Gemini all use RAG selectively when queries require fresh information.
Refresh cadence How often you update your content to maintain freshness. For high-priority GEO pages, a meaningful update every 7–14 days is a reasonable baseline. Quarterly is typically too slow for fast-moving categories.
Robot (or AI crawler) The automated agent an AI engine uses to crawl and index web content. ChatGPT's is ChatGPT-User (real-time browsing) and GPTBot (training crawl). Anthropic uses ClaudeBot. Perplexity uses PerplexityBot. Don't block these unless you have a specific reason to — blocking them removes you from AI search.
S
Schema.org A vocabulary of structured data types (Product, Offer, Review, FAQPage, etc.) maintained collaboratively by Google, Microsoft, Yandex, and Yahoo. The primary mechanism for exposing structured facts to AI engines. Schema.org markup, rendered as JSON-LD, is critical for GEO.
Server-side rendering (SSR) Rendering HTML on the server before sending it to the browser, rather than relying on JavaScript to inject content client-side. SSR is functionally required for GEO because many AI crawlers don't execute JavaScript reliably. Content rendered only via JS is often invisible to AI.
Share of voice (in AI search) Your brand's percentage of all relevant brand mentions across a defined set of AI queries. If competitors are mentioned 7 times across 100 queries and you're mentioned 3 times, your share of voice is 30% (3 of 10 total mentions).
Source rate The percentage of AI responses that use your domain as a source (cited as reference), regardless of whether your brand is named in the answer. Source-only links are leading indicators of brand citations — AI engines that trust your content as a reference are closer to citing your brand.
Structured data Machine-readable information embedded in HTML, typically using Schema.org vocabulary, that tells AI engines and search crawlers about the content of a page in a typed, parseable way. Distinct from prose, which requires interpretation.
T
Token The unit of text an LLM processes. Roughly equivalent to a syllable. Models have token limits both for input (the prompt and retrieved context) and output (the generated response). Token budgets influence how much retrieved content an AI engine can actually use when answering.
Topical authority A loose measure of how much an AI engine considers your domain to be an authoritative source on a specific topic. Built through depth of content (multiple substantive pieces on related topics), original research, and third-party validation.
V
Vector search The technique of searching by semantic similarity (via embeddings) rather than literal keyword matching. Most modern AI retrieval systems use vector search either alone or alongside traditional keyword search.
Y
YMYL (Your Money or Your Life) A Google search quality framework, adopted in spirit by AI engines, identifying topics where misinformation could harm the user — health, finance, safety, legal. AI engines apply stricter trust filters to YMYL content. Supplements and wellness fall partially in this category; cosmetics fall closer to the boundary.
Z
Zero-click search A search where the user's question is answered directly in the results page (or the AI response) without requiring a click. AI search dramatically increases zero-click outcomes, which is why brand mentions matter more than clicks in modern GEO measurement.