Reference

AI Glossary: Every Term Explained in Plain English

AI writing gets jargon-heavy fast. This is a plain-English reference for the terms that actually come up — grouped by category, no unnecessary complexity, cross-linked to the fuller guides where one exists.

Key takeaways

Most AI jargon describes a small number of underlying ideas — how models are trained, how you instruct them, and how they're connected to other systems.
Understanding "hallucination" and "context window" alone explains most of the practical limitations people run into.
GEO (Generative Engine Optimization) is the AI-search counterpart to traditional SEO — same goal, different audience (AI assistants instead of search rankings).

Core AI Concepts

Artificial Intelligence (AI)

Software that performs tasks — understanding language, recognizing patterns, generating content — that would otherwise require human judgment. Today's most common consumer form is the large language model.

Large Language Model (LLM)

A model trained on huge amounts of text to predict the next piece of text given what came before. ChatGPT, Claude, and Gemini are all built on LLMs. It generates language; it doesn't look facts up in a database unless a separate tool (like search) is added.

Generative AI

AI that creates new content — text, images, audio, video, code — rather than just classifying or analyzing existing content. LLMs, image generators, and voice-cloning tools all fall under this.

Hallucination

When an AI model states something confidently and fluently that is factually wrong. It happens because the model is generating plausible-sounding text, not retrieving verified facts — always worth double-checking specific dates, numbers, or citations an AI gives you.

Training Data

The text (or images, audio) a model learned from before it was released. A model's knowledge has a cutoff date tied to when its training data was collected — it generally can't know about things that happened after that, unless it has live search access.

Multimodal

An AI system that can handle more than one type of input or output — text and images, or text and voice — rather than text only.

Prompting Terms

Prompt

The instruction or question you give an AI model. A prompt can be one line or a detailed brief with context, format, and constraints — see Prompt Engineering below.

Prompt Engineering

The practice of structuring instructions — role, context, format, examples, constraints — to reliably get a specific kind of output from an AI model, especially for repeatable tasks. See the full guide at /what-is-prompt-engineering.

System Prompt

A standing instruction set at the start of a conversation that shapes how the AI behaves for everything after it — used to set a persona, tone, or rules once instead of repeating them every message.

Few-Shot Prompting

Giving the AI a few examples of the exact output format or style you want before asking for the real one, so it pattern-matches to your examples rather than guessing your intent from a plain instruction.

Chain-of-Thought Prompting

Asking a model to reason through a problem step by step before giving a final answer, which tends to produce more accurate results on multi-step or logic-heavy questions than asking for the answer directly.

Context Window

The amount of text an AI model can "see" at once — the current conversation, any pasted documents, and its own reply all count against this limit. A larger context window means it can hold more of a long document or conversation in mind at once.

Temperature

A setting (mostly relevant via API, not the chat interface) controlling how random or predictable a model's output is — lower is more consistent and literal, higher is more varied and creative.

Technical Terms

Token

The basic unit an AI model processes text in — roughly a word or part of a word. Model pricing and context-window limits are usually measured in tokens, not characters or words.

Fine-Tuning

Additional training on a specific, narrower dataset to make a general model better at a particular task or style — distinct from prompting, which shapes behavior without changing the underlying model.

RAG (Retrieval-Augmented Generation)

A technique where a model retrieves relevant real documents (a knowledge base, a website) before generating its answer, so the response is grounded in that retrieved content instead of purely the model's training data — one of the main ways hallucination gets reduced in real products.

AI Agent

An AI system that can take multiple steps toward a goal with limited human input per step — deciding what to do next, using tools, and adjusting based on results — rather than producing one response to one prompt. See the full guide at /what-are-ai-agents.

MCP (Model Context Protocol)

An open standard that lets AI models connect to external tools and data sources in a consistent way — for example, letting an AI assistant read from or write to a specific app's data through a standardized connection, rather than a one-off custom integration for every tool.

API (Application Programming Interface)

The way developers connect their own software directly to an AI model, instead of using its chat interface — this is how AI features get built into other products.

AI Business Terms

AI Workflow

A sequence of AI-assisted steps chained together to complete a larger task — for example, research → outline → draft → edit — rather than one prompt doing everything at once.

Productized Service

A service sold with a fixed scope and fixed price (e.g., "a 30-day content calendar, done for you") instead of open-ended custom work billed by the hour — generally easier to price, sell, and deliver consistently.

Credit-Based Pricing

A pricing model (used by AIVX Labs and similar platforms) where a membership includes a monthly allowance of credits, and each action (running a tool, generating an asset) spends a set number of credits rather than being billed individually.

GEO & AI Search Terms

GEO (Generative Engine Optimization)

The practice of structuring a website so AI assistants (ChatGPT, Claude, Gemini, Perplexity) are more likely to cite or recommend it when answering a relevant question — the AI-era counterpart to traditional SEO, which optimizes for search engine rankings instead.

Entity (SEO/GEO sense)

A distinctly named thing — a company, product, or concept — that a search engine or AI system can recognize and associate facts with. Consistent naming and structured data help a site become clearly identified as an authority on specific entities.

Structured Data / Schema Markup

Machine-readable metadata embedded in a webpage (typically JSON-LD) describing what the page is — an article, a product, an FAQ — in a standardized format search engines and AI crawlers can parse directly, rather than inferring it from visual layout.

Frequently asked questions

What does LLM stand for?

Large Language Model — the underlying technology behind ChatGPT, Claude, Gemini, and similar AI assistants.

What is the difference between AI and generative AI?

AI is the broad category of software that performs tasks requiring human-like judgment. Generative AI is the subset that creates new content — text, images, audio — rather than only classifying or analyzing existing content.

What is GEO in the context of AI, not location?

Generative Engine Optimization — structuring a website so AI assistants like ChatGPT, Claude, and Perplexity are more likely to cite or recommend it, the same way traditional SEO optimizes for search engine rankings.