ChatGPT & Prompting Terms Glossary

AI Agent

An AI agent is an AI that can take action on your behalf — not just answer questions, but actually do things. It can browse the web, send emails, complete multi-step tasks, and make decisions with minimal input from you.

💡 It’s less like a calculator and more like a capable assistant you can brief once and then let run with the task.

AI Assistant

An AI assistant is a tool that helps you get things done — answering questions, drafting content, summarising documents, or saving you time. ChatGPT, Claude, and Gemini are all AI assistants.

💡 Think of it as a knowledgeable colleague available 24/7 who never gets tired and never judges your questions.

AI Automation

AI automation uses artificial intelligence to handle tasks that normally need human thinking — adapting, making decisions, and responding to different situations — not just following rigid rules.

💡 A chatbot that reads a customer complaint and drafts a personalised reply without you lifting a finger is AI automation.

AI Chatbot

An AI chatbot is a conversational tool you can talk to through text (or voice). It can answer questions, guide you through tasks, and handle common requests — without a human on the other end.

💡 The chat bubble on a company’s website that answers your questions at midnight is usually an AI chatbot.

AI Citation

An AI citation is when an AI tool references the source it pulled information from while generating an answer. It’s AI’s way of showing its work.

💡 If an AI answer includes a link or says “according to [website]” — that’s a citation. A good sign the answer is grounded in real sources.

AI Content Detection

AI content detection tools try to identify whether a piece of writing was generated by AI or written by a human. They look for patterns in sentence structure and word choice that AI tends to repeat.

💡 These tools aren’t perfect — they can flag human writing as AI-generated and miss actual AI content. Use them as a guide, not a verdict.

AI Copilot

An AI copilot is an AI assistant built directly into the tools you already use — like Microsoft Word or your email — helping you work faster without switching apps.

💡 Microsoft Copilot inside Word, helping you draft and edit without leaving the document — that’s an AI copilot.

AI Hallucination

AI hallucination is when an AI makes something up — and sounds completely confident doing it. It might invent a statistic, a quote, or even a person that doesn’t exist. It’s not lying; it’s a genuine limitation of how these systems work.

💡 Always double-check any facts, figures, or names that AI gives you — especially anything you plan to publish or share.

AI Image Generation

AI image generation creates brand new images from a text description — no design skills required. You describe what you want, and the AI draws it.

💡 “A friendly robot drinking coffee at a Cape Town café, watercolour style” — tools like Midjourney or DALL·E will turn that into an image in seconds.

AI Literacy

AI literacy is understanding how AI works well enough to use it wisely — knowing what it can and can’t do, when to trust it, and when to question it. You don’t need to be technical to be AI literate.

💡 Reading this glossary is already a step toward AI literacy. You’re doing it right now.

AI Memory

AI memory allows some AI tools to remember information from past conversations — your name, your preferences, your previous requests — so it doesn’t start from scratch every time.

💡 If ChatGPT remembers you’re a small business owner and tailors its replies accordingly, that’s AI memory at work.

AI Model

An AI model is the trained system behind an AI tool — the brain that processes your input and generates a response. When you use Claude or ChatGPT, you’re interacting with an AI model.

💡 Different models have different strengths. Some are better at writing, others at coding or analysis.

AI Overview

An AI Overview is the summary box that appears at the top of some Google searches — a direct, AI-generated answer to your question, pulled from multiple websites.

💡 You’ve probably seen these already. They appear above the regular search results and give you the short answer before you’ve clicked anything.

AI Productivity Tools

AI productivity tools are apps and software that use AI to help you save time — drafting emails, summarising documents, automating tasks, organising notes, and more.

💡 ChatGPT, Claude, Notion AI, and Grammarly are all examples of AI productivity tools.

AI Search

AI search gives you a direct answer instead of a list of links to scroll through. It reads multiple sources, pulls out the relevant information, and presents it in plain language.

💡 Ask Perplexity or Google’s AI Mode a question and get a summarised answer with sources — that’s AI search.

AI Video Generation

AI video generation creates or edits videos automatically using text prompts, scripts, or existing clips — no video editing experience needed.

💡 Tools like Sora or Runway can turn a written script into a short video clip with visuals, voiceover, and music.

AI Voice Generation

AI voice generation converts written text into realistic spoken audio — a human-sounding voice reading your content aloud, without needing a real person in a recording studio.

💡 Podcast creators use this to turn blog posts into audio episodes. Customer service teams use it for phone systems.

AI Workflow

An AI workflow is a sequence of connected tasks where AI tools work together to get something done — like drafting, reviewing, and scheduling content all in one connected chain.

💡 AI writes your social post → another tool resizes the image → a scheduler posts it automatically. That chain is an AI workflow.

Atomic Answer

An atomic answer is a short, self-contained response that answers one question completely — no extra context needed. AI systems love these because they’re easy to extract and use.

💡 “What is a prompt? A prompt is the instruction you give to an AI tool.” Short, complete, standalone. That’s an atomic answer.

Bias in AI

AI bias happens when an AI produces unfair or skewed results because the data it learned from was unbalanced. If an AI mostly learned from one type of person or perspective, it may perform worse for everyone else.

💡 An AI hiring tool trained mostly on male CVs might unfairly score female candidates lower. That’s AI bias — and it’s a real problem worth knowing about.

Chain-of-Thought Prompting

Chain-of-thought prompting is when you ask an AI to think through a problem step by step before giving its final answer — rather than jumping straight to a conclusion. It usually leads to better, more reasoned responses.

💡 Add “think through this step by step” to your prompt and watch the quality of the answer improve — especially for complex questions.

ChatGPT

ChatGPT is an AI chatbot created by OpenAI. It can answer questions, write content, summarise information, help with research, brainstorm ideas, and assist with a wide range of tasks through a simple chat interface.

💡 It’s one of the most widely used AI tools in the world — and a good starting point if you’re new to AI.

Context Window

The context window is the AI’s short-term memory. It can only hold so much information at once during a conversation. If a chat gets very long, earlier parts can “fall off” and the AI may forget what was said at the start.

💡 For long projects, paste in the key details again partway through — just to keep the AI on track.

Conversational AI

Conversational AI is the broader term for AI systems designed to hold natural back-and-forth conversations with humans — understanding context, follow-up questions, and nuance, not just single commands.

💡 The difference between an old-school phone menu (“press 1 for billing”) and a modern AI that understands what you’re actually asking — that’s conversational AI.

Conversational Search

Conversational search lets you search the way you actually talk — asking follow-up questions, refining your request, and having a back-and-forth rather than typing a brand new search each time.

💡 “Best CRM for small business?” → “Which works best with Gmail?” → “How much does that one cost?” — that’s conversational search.

Custom GPT

A Custom GPT is a personalised version of ChatGPT that you configure with specific instructions, knowledge, files, and behaviours — built for a particular task or business need.

💡 A marketing agency might build a Custom GPT trained on their brand voice and client briefs, so every team member gets consistent AI output.

Dataset

A dataset is an organised collection of information used to teach or test an AI. It could be thousands of photos, customer records, product reviews — any structured set of information an AI can learn from.

💡 Think of it as the AI’s study pack — everything it needs to prepare for the task.

Embeddings

Embeddings are how AI understands that words with similar meanings belong together — even if they look completely different. It converts words and phrases into numbers that capture meaning and relationships.

💡 It’s why an AI knows “dog” and “puppy” are related, or that “happy” and “joyful” mean roughly the same thing — without being told explicitly.

Ethical AI

Ethical AI is about making sure AI is built and used in ways that are fair, honest, and safe — that it doesn’t discriminate, mislead people, or cause harm.

💡 “Could this AI make unfair decisions about job applications?” or “Is my data being used responsibly?” — these are ethical AI questions worth asking.

Few-Shot Prompting

Few-shot prompting means giving the AI a couple of examples of what you want before asking it to do the task. Those examples help it understand exactly the format, tone, or style you’re after.

💡 Show it two example product descriptions in your brand voice, then ask it to write a third. It’ll match your style far better than if you just asked cold.

Fine-Tuning

Fine-tuning is when an existing AI model gets extra training on a specific type of content — legal documents, customer service scripts, medical notes — to make it much better at that particular area.

💡 Think of it like a general-purpose chef who then does a specialised course in French cuisine. Same base skills, sharper focus.

Generative AI

Generative AI is AI that creates new things — text, images, music, video, or code — from scratch, based on a prompt you give it. It doesn’t copy and paste. It generates something new every time.

💡 ChatGPT writing you a marketing email, or Midjourney drawing a logo from your description — both are generative AI.

Large Language Model (LLM)

An LLM is an AI system trained on enormous amounts of text — books, websites, articles — so it can understand and generate human language. It’s the engine behind most AI chatbots and writing tools.

💡 ChatGPT, Claude, Gemini, and Copilot are all examples of large language models.

Multimodal AI

Multimodal AI can handle more than just text — it can understand and create images, audio, video, and text all at once. “Multi” means many; “modal” means types of input or output.

💡 Upload a photo to Claude or ChatGPT and ask it to describe what’s in it — that’s multimodal AI in action.

Natural Language Processing (NLP)

NLP is the part of AI that helps computers understand human language — not just the words, but the meaning and intent behind them. It’s what allows AI to tell the difference between a question and a command.

💡 When you talk to a voice assistant and it actually understands what you meant (not just what you said), that’s NLP doing the work.

Passage Retrieval

Passage retrieval is when an AI scans through a large amount of content and pulls out the most relevant sections to answer your question — rather than reading everything word by word.

💡 It’s like doing a ctrl+F search, but the AI actually understands meaning, not just matching keywords.

Prompt

A prompt is simply what you type to an AI — your question, request, or instruction. The quality of your prompt has a huge impact on the quality of what comes back.

💡 “Write me a blog post” is a prompt. So is “Write a 500-word blog post for small business owners about saving time with AI — keep it friendly and jargon-free.” The second one will get a much better result.

Prompt Engineering

Prompt engineering is the skill of writing better prompts to get better results from AI. It’s not about coding — it’s about being clear, specific, and giving the AI the right context to do its job well.

💡 Anyone can learn this. The better your instructions, the better the output. Think of it as learning how to give a really good brief.

Prompt Library

A prompt library is a collection of ready-made prompts you can reuse for different tasks — so you’re not starting from scratch every time you need AI to do something.

💡 Think of it like a swipe file for AI instructions. Build one up over time and your productivity will compound.

Prompt Template

A prompt template is a reusable prompt structure with gaps you fill in — so you can produce consistent, quality outputs for repeated tasks without rewriting your instructions every time.

💡 “Write a [tone] email to [audience] about [topic], no longer than [length].” Fill in the blanks, get a great draft every time.

Query Fan-Out

Query fan-out is when an AI breaks your single question into several smaller searches behind the scenes before pulling together a complete answer. You ask one thing; it investigates multiple angles.

💡 You ask “What’s the best AI tool for small business marketing?” and the AI quietly searches five related questions before giving you one clear answer.

Retrievability

Retrievability is how easy it is for AI systems to find and extract information from your content. Well-structured, clearly written content is far more retrievable than dense, jargon-heavy pages.

💡 Short paragraphs, clear headings, and direct answers all improve retrievability — for both humans and AI.

Retrieval-Augmented Generation (RAG)

RAG is when an AI looks something up before it answers — like checking a book instead of guessing from memory. It pulls in real, up-to-date information from external sources before generating its response.

💡 If an AI can read your company documents and answer questions based on them, it’s probably using RAG.

Role Prompting

Role prompting is when you tell the AI to act as a specific type of expert before asking your question. It helps the AI adjust its tone, depth, and focus to match what you actually need.

💡 “Act as a marketing strategist for small businesses” before asking for campaign ideas will get you a far more targeted response than asking cold.

Semantic Search

Semantic search understands what you mean, not just what you typed. Instead of hunting for exact keyword matches, it grasps the intent behind your question and finds more relevant results.

💡 Search “good shoes for sore feet” and get results for orthopedic footwear — even though you didn’t use that word. That’s semantic search.

Speech Recognition

Speech recognition is AI that converts spoken words into text or commands. It’s what powers voice assistants, meeting transcriptions, and live captioning.

💡 Dictating a message on your phone, or asking Siri a question — both use speech recognition.

System Prompt

A system prompt is a hidden set of instructions that shapes how an AI behaves before you even start talking to it — setting its tone, rules, and focus. You usually don’t see it, but it’s always there.

💡 When an AI chatbot on a website only talks about that company’s products, a system prompt is telling it to stay on topic.

Temperature

Temperature is a setting that controls how creative or predictable an AI’s responses are. Low temperature = more focused and consistent. High temperature = more varied and creative.

💡 For factual tasks like summarising a report, you want low temperature. For brainstorming campaign ideas, turn it up.

Token

A token is a tiny chunk of text — roughly one word, or sometimes just part of a word — that the AI reads and processes. AI doesn’t read sentences the way we do; it processes token by token.

💡 This matters mainly for understanding why very long conversations or documents can hit limits or slow things down.

Training Data

Training data is the information an AI learned from before you ever spoke to it — millions of books, articles, websites, and more. The better the training data, the smarter and more accurate the AI tends to be.

💡 Just like a student learns from textbooks, an AI learns from its training data. Rubbish in, rubbish out.

User Prompt

A user prompt is the message you type into an AI tool — your question, instruction, or request. Every conversation starts with a user prompt.

💡 The more specific and clear your user prompt, the better the AI’s response will be. Vague in, vague out.

Zero-Shot Prompting

Zero-shot prompting means asking the AI to complete a task with no examples — just your instruction. It relies entirely on the AI’s existing training to figure out what you want.

💡 “Write me a product description for a handmade soy candle” is zero-shot. No examples given — the AI figures out the