Back homeAI & prompting,
AI & prompting,
in plain English.
Every term you'll bump into when learning AI prompting — defined simply, with no assumed background.
- Prompt
- The instruction you give an AI tool. Anything from a one-line question to a structured paragraph telling the AI who to be, what to do, and how to format the answer.
- Prompt engineering
- A fancy phrase for 'writing better prompts'. It just means thinking about your prompt before you hit send — being clear about role, task, context, and the output you want.
- Prompting framework
- A short recipe for writing prompts. Examples: RTF (Role · Task · Format), RACE (Role · Action · Context · Expectation). They give you slots to fill in so you don't forget anything important.
- LLM (Large Language Model)
- The kind of AI behind ChatGPT, Gemini, and Claude. It's a system that has read enormous amounts of text and learned to predict useful, sensible next words.
- ChatGPT
- OpenAI's chat-based AI assistant. Free and paid tiers. One of the two most popular tools we write prompts for.
- Gemini
- Google's chat-based AI assistant. Free and paid tiers. The other tool we write prompts for. Same prompting frameworks work in both.
- Claude
- Anthropic's chat-based AI assistant. Also responds well to the same prompting frameworks we cover.
- Token
- A chunk of text the AI reads — usually a short word or part of a word. Most prompts are well under any free-tier token limit, so beginners rarely need to worry about it.
- Hallucination
- When an AI confidently says something that isn't true. The fix is simple: sanity-check anything important against a reliable source.
- Context window
- How much text the AI can 'remember' in one conversation. If a long chat starts going off the rails, start a new chat.
- System prompt
- Hidden instructions that shape how an AI assistant behaves before you say anything. As a regular user you don't usually set this — your first message is enough.
- Few-shot prompting
- Showing the AI 1–3 examples of what 'good' looks like, then asking it to do the same for your case. Hugely effective for tone and style.
- Zero-shot prompting
- Asking the AI to do something with no examples — just a clear instruction. Most everyday prompts are zero-shot.
- Chain-of-thought
- Asking the AI to 'think step by step' before answering. Often improves reasoning on tricky questions.
- Iteration
- Replying to the AI's answer to improve it — 'shorter', 'more casual', 'add an example'. The single most underused beginner skill.
Daily by email