What the words actually mean — and why each one matters for whether AI can find your business.
There's a lot of vocabulary in this space, and most of it got invented in the last two years. Some of it is genuinely useful. Some of it is the same old idea wearing a new jacket. This is the working list — what each term means, and the part that actually matters for a business trying to show up in AI answers.
Core AI Search & GEO
Generative Engine Optimization (GEO)
The practice of structuring your business information so AI engines can find it, understand it, trust it, and repeat it back when someone asks for a recommendation.
The name is new; the discipline isn't entirely. What's new is the target — you're no longer optimizing for a position on a page. You're optimizing to be named in a sentence.
AI Search
Any search experience where a generative model answers the question directly instead of returning a list of links. ChatGPT, Perplexity, Google's AI Overviews, Gemini, Claude.
The thing to sit with: there's no page two. There's no scrolling down to find you at position seven. You're in the answer or you're not.
Answer Engine Optimization (AEO)
A narrower slice of GEO focused on formatting content so it can be lifted directly as an answer — clear questions, direct responses, structured sections.
Mostly a formatting discipline. Useful, but it's the easy half. Getting your content extractable is straightforward; getting an engine to trust you enough to use it is the hard part.
Large Language Model (LLM)
The type of AI model underneath ChatGPT, Claude, Gemini, and the rest. Trained on enormous amounts of text, then used to generate, summarize, and reason with language.
You don't need to understand how they work. You need to understand what they need from you: clear, consistent, verifiable information about what your business is.
Answer Engine
A search or assistant product that returns a synthesized answer rather than a list of sources.
Plain-English alternative to "generative engine" when "LLM" makes eyes glaze over.
AI Overview
Google's AI-generated summary that appears above traditional search results for many queries.
This is where most business owners first notice something changed — their ranking didn't move, but the traffic did.
AI Mode
Google's dedicated conversational search interface, separate from AI Overviews embedded in regular results.
Different surface, different citation behavior. Worth testing separately rather than assuming one predicts the other.
Zero-Click Search
A search where the person gets their answer without clicking through to any website.
The uncomfortable version of this: your content can be the source of an answer that never sends you a visitor. Being useful and being visited have come apart.
Prompt
The question or instruction someone types into an AI tool.
Worth internalizing that prompts are not keywords. People type full sentences with context and constraints — "a dentist near downtown who takes my insurance and has Saturday hours" — and the engine answers all of it at once.
Prompt Set
A defined group of prompts used repeatedly to test whether AI engines mention a business. Usually spans branded, unbranded, local, and comparison questions.
This is what turns "I asked ChatGPT once and it didn't know me" into something you can actually measure over time.
Branded Prompt
A prompt that names your business directly — "What does [company] do?" or "Is [company] any good?"
Start here in any audit. If an engine can't describe you accurately when handed your own name, nothing downstream is going to work.
Unbranded Prompt
A prompt that describes what someone wants without naming a company — "best roofer in Tucson."
This is the one that matters commercially. Branded prompts tell you whether AI knows you. Unbranded prompts tell you whether AI recommends you.
Recommendation Query
A prompt asking the AI to suggest a business, product, or provider.
These decide shortlists before anyone visits a website. By the time someone lands on your page, the comparison may already be over.
Comparison Query
A prompt asking AI to weigh options against each other — "X vs Y," "alternatives to Z."
Late-stage, high-intent, and largely out of your hands unless you've published something the engine can use to describe you accurately.
Entities & How AI Understands Businesses
Entity
A thing a machine can recognize as a distinct object — a business, a person, a place, a product.
This is the unit AI engines actually deal in. They don't recommend web pages; they recommend entities. Which is why a business can have a perfectly good website and still not exist as far as an AI is concerned.
Entity Trust
How confident a system is that a business is real, consistent, and credible enough to name.
Built slowly out of boring things: matching details across platforms, reviews that accumulate, other sites mentioning you. There's no shortcut, which is either frustrating or reassuring depending on your position.
Entity Home
The single page that most clearly states who you are, what you do, where you operate, and why you're credible. Usually a homepage or a thorough About page.
Most sites don't have one. They have a homepage full of adjectives and a contact form. If a machine had to summarize your business from one page, which page would you want it to read?
Disambiguation
The process of an AI figuring out which entity you actually mean when several share a name or a similar description.
If there are four businesses in your city with names close to yours, the engine has to pick. Structured, consistent, cross-referenced information is how it picks correctly.
Knowledge Graph
A structured database of entities and the relationships between them — who works where, what's located in what city, which product belongs to which brand.
Search engines have been building these for over a decade. AI answers lean on them heavily. Being in one is more valuable than ranking for a keyword.
Entity Grounding
Anchoring your business to authoritative external references — linking your identity to established databases and profiles so the engine has something to verify against.
The difference between claiming you're a real company and being verifiably one.
Semantic Search
Search based on meaning and relationships rather than literal keyword matching.
The practical implication: writing naturally about what you actually do beats stuffing a page with the phrase you hope someone types.
The Technical Layer
Structured Data
Information formatted in a standardized, machine-readable way so systems don't have to guess what they're looking at.
The gap between "our hours are on the website somewhere" and "our hours are stated in a format every machine on the internet already knows how to read."
Schema Markup
The specific vocabulary (from Schema.org) used to label business information as structured data — hours, address, services, reviews, products.
It won't make an AI recommend you. It removes the excuse for an AI to misunderstand you. Necessary, not sufficient.
JSON-LD
The code format used to add schema markup to a page. It sits in the page's code without affecting what visitors see.
The format Google explicitly prefers. If you're adding schema, add it this way.
Nested Schema
Schema where related entities are embedded inside each other rather than listed separately — a person inside an organization, a product inside a brand.
Flat schema hands the machine a pile of facts. Nested schema hands it the relationships between them. The second one requires less guessing.
LocalBusiness Schema
The schema type for physical businesses — address, hours, service area, contact details, geographic coordinates.
If you have a location customers visit, this is the single highest-value piece of markup on your site.
Organization Schema
The schema type establishing a company's core identity — name, logo, contact info, official profiles.
Where you tell machines, unambiguously, who you are. Often the first thing missing.
FAQPage Schema
Markup that labels question-and-answer pairs on a page as exactly that.
AI engines are answering questions. Content already shaped like a question and an answer takes less work to use. This is one of the more direct wins available.
Product Schema
Markup describing a product — price, availability, specifications, ratings, identifiers.
For ecommerce this is the whole ballgame. AI shopping assistants read structured product data, not your marketing copy.
sameAs
A schema property linking your entity to its profiles elsewhere — social accounts, directory listings, authoritative references.
A small, cheap, frequently skipped signal that says "these accounts and this business are the same thing."
robots.txt
A file telling crawlers which parts of your site they may access.
Worth checking. A surprising number of businesses are invisible to AI crawlers because of a rule someone added years ago for a reason nobody remembers.
llms.txt
A proposed plain-text file giving AI systems a clean, structured map of a site's most important content.
Genuinely useful for reducing how much work a model does to understand your site. But it is a proposal, not a standard — adoption varies by engine, and anyone presenting it as a required ranking factor is getting ahead of the facts.
Crawlability
Whether automated systems can actually reach and read your content.
Content locked behind interactive elements, loaded only by scripts, or hidden behind a login may not exist as far as a crawler is concerned. Whatever matters most should be in the page itself.
Server-Side Rendering
Delivering fully-formed HTML from the server rather than assembling the page in the visitor's browser.
Humans don't notice the difference. Machines do — and the ones that don't wait around for scripts to finish simply see an empty page.
Retrieval-Augmented Generation (RAG)
The process where an AI looks up real information before answering, rather than relying only on what it memorized during training.
This is why current, accurate, findable information about your business matters. The model isn't recalling you — it's looking you up in the moment.
Retrieval
The lookup step itself — finding relevant sources before generating an answer.
If your information isn't retrievable, everything else is academic.
Trust, Reviews & Third-Party Signals
NAP
Name, Address, Phone — the core identity details of a local business.
The most boring item on this list and one of the most consequential. Three fields, and getting them to match everywhere is harder than it sounds.
NAP Consistency
Those details matching exactly across every platform where your business appears.
A suite number missing on one listing and present on another is enough to make a machine less certain you're one business. Confusion doesn't get you penalized — it gets you skipped.
Citation (local)
Any mention of your business name, address, and phone on another website — directories, chambers, industry lists.
Old-fashioned, unglamorous, and still functioning as corroboration. The more independent places that agree on your details, the more confident a system can be.
Citation (AI)
A source an AI engine references or links when generating an answer.
Confusingly, the same word means two different things in this field. Local citations are listings. AI citations are being used as a source. Both matter; they're not the same thing.
Google Business Profile
The business listing that appears in Google Search and Maps.
For local businesses this is frequently the single most influential asset — more than the website. It's also the one most often left half-finished.
Review Velocity
The rate at which new reviews come in over time.
A hundred reviews from three years ago reads differently than twenty from the last three months. Recency is doing work here that volume alone doesn't.
Review Sentiment
The overall tone of what people say about you, not just the star rating.
Ratings are a number. Sentiment is the actual language, and language is what a model reads. What people say about you specifically is what gets repeated.
Third-Party Proof
Anything credible said about your business somewhere you don't control — press, industry directories, review platforms, mentions on other sites.
Your own website is a claim. Everything else is evidence. Engines weigh evidence more heavily, for the obvious reason.
E-E-A-T
Experience, Expertise, Authoritativeness, Trustworthiness — Google's framework for assessing content credibility.
An acronym doing a lot of work, but the underlying idea is sound: can a machine tell that a real, qualified person or organization is behind this?
YMYL
"Your Money or Your Life" — topics touching health, finance, safety, or legal matters, held to stricter credibility standards.
If you're in one of these fields, the bar for demonstrating credibility is higher, and the shortcuts that work elsewhere don't work here.
Content
Answer-First Content
Writing that states the answer immediately, then explains — rather than building up to it.
Contrary to how a lot of marketing copy is written. The payoff sentence goes first, because an engine extracting an answer takes the clearest one available.
Fact Density
How many concrete, verifiable specifics a piece of content contains per unit of text — numbers, dates, named things.
One of the more actionable findings in the research: specifics get extracted, adjectives get skipped. "Serving the area since 2011 across four counties" does work that "trusted local experts" cannot.
Topical Authority
Being recognized as a credible source across a subject area, not just one page.
Built by covering a topic thoroughly rather than publishing one page and hoping. Slow, and difficult to fake.
Comparison Content
Pages that honestly compare options, approaches, or providers.
People ask AI comparison questions constantly. If nothing credible exists describing how you compare, the engine assembles an answer from whatever it can find — which may be a competitor's version.
Thin Content
Pages with little substance, usually produced in volume to cover keyword variations.
The strategy that used to work and now actively hurts. Volume without substance is the specific thing recent search updates have been built to catch.
Scaled Content Abuse
Mass-producing low-value pages, AI-generated or otherwise, primarily to manipulate search visibility.
Worth being precise: the problem isn't that content was AI-assisted. The problem is content that exists for machines rather than people. Those are different things, and only one gets penalized.
Measurement
AI Visibility
Whether, and how often, your business appears in AI-generated answers across engines.
The metric that replaces "what position do we rank." Harder to measure, and it varies by engine, phrasing, and week.
Share of Voice
How often you appear compared to competitors across the same set of prompts.
More useful than your own number in isolation. "We appear 20% of the time" means something different if the leader is at 25% versus 80%.
Mention vs. Citation
Being named in an answer versus being used and linked as a source.
Two different outcomes with different value. A mention influences the reader. A citation influences the reader and can send them to you.
Prompt Volatility
How much AI answers change across small variations in wording, or across repeated identical asks.
The reason single spot-checks mislead. Asking once and seeing your name doesn't mean you're visible; it means you were visible that time.
Cross-Engine Variance
The differences in what each AI engine says about the same business.
They're trained differently, retrieve differently, and weight sources differently. Being strong in one is not evidence of being strong in the others — which is exactly why checking only one gives false comfort.
Traditional SEO Terms That Still Matter
Indexing
Whether a search engine has stored your page and can return it.
Still foundational. Content that was never indexed can't be retrieved, cited, or recommended by anything.
Canonical URL
The tag marking which version of a page is the authoritative one when duplicates exist.
Prevents your own pages from competing with each other and muddying which one is real.
Backlinks
Links from other websites to yours.
Still a signal, but no longer the dominant one. Recent research suggests plain unlinked brand mentions now correlate more strongly with AI citation than links do — which is a meaningful shift in what "authority" means.
Core Web Vitals
Google's page-experience measurements — loading speed, interactivity, visual stability.
Matters for human visitors and for ranking. Less directly relevant to AI citation, but a slow, broken site isn't helping anything.
Search Intent
What someone is actually trying to accomplish with a query — learning, comparing, or buying.
Unchanged as a concept, more visible in practice. AI answers make intent explicit because people write full sentences instead of three keywords.
This list will keep changing. That's not a disclaimer — it's the actual condition of working in this field right now. Terms that seem essential this quarter may be footnotes by next year, and things nobody has named yet will matter more than half of what's here.