Most AI words sound complicated and mean simple things. Here are the ten you’ll hear most often, in plain English. A prompt is what you type. A model is the AI itself. A hallucination is when AI makes something up. A token is a chunk of text. The rest are below. Save this and you’ll never have to fake-nod through an AI conversation again.
1. Prompt
What you type to tell the AI what you want. Your message, your request. “Write me a grocery list” is a prompt.
2. Model
The AI itself, the “brain” behind the tool. ChatGPT, Claude, and Gemini are each built on a model. When someone says “which model are you using,” they mean which AI.
3. Generative AI
AI that creates things, like text, images, or audio. The “generative” just means it generates new content rather than only sorting or searching. ChatGPT is generative AI.
4. Hallucination
When AI confidently makes something up, a fake fact, a made-up quote, a wrong number. It’s the main reason to double-check anything important. The AI isn’t lying on purpose, it’s predicting an answer and sometimes predicts wrong.
5. Token
The small chunks of words AI reads and writes in. A token is roughly part of a word. It mostly matters because it’s how AI measures length. You rarely need to think about it.
6. Prompt engineering
A fancy term for asking better questions to get better answers. That’s the whole thing. You’re already doing it when you add detail to a request.
7. LLM (Large Language Model)
The type of AI behind tools like ChatGPT. “Large” because it learned from a huge amount of text, “language model” because it works with words. When you hear LLM, think “the text-based AI.”
8. Training
How an AI learned in the first place, by reading enormous amounts of text and finding patterns. Training happened before you ever opened the tool. It’s why AI “knows” things.
9. Context
Everything the AI is keeping in mind during your conversation, including what you said earlier. Giving it good context, like who a message is for, leads to much better answers.
10. Agent
An AI that can take actions for you, not just chat, like booking something or working through a multi-step task on its own. Agents are newer and growing fast. For now, most everyday use is still chat.
Five more you’ll bump into
Once the first ten feel familiar, these come up often enough to be worth a quick definition.
Chatbot. A program you talk to in a chat window. ChatGPT is a chatbot. The word just describes the back-and-forth format.
Multimodal. AI that handles more than text, like images, audio, or video. A multimodal AI can look at a photo you upload and describe it, not just read your words.
Open source. AI that’s released freely for anyone to use or build on, versus “closed” AI owned by one company. You’ll see this in news about new models.
Fine-tuning. Taking a general AI and training it a bit more on a specific topic so it gets better at one job. You don’t need to do this, but you’ll hear the term.
Bias. AI learns from human-made text, so it can pick up human biases and blind spots. It’s a real limitation, and a reason your own judgment still matters.
How these fit together
If you want the big picture, here’s how the words connect. A model is the underlying AI brain, and the chat-based kind is called an LLM. It got smart through training on huge amounts of text. You interact with it through a chatbot, where you send a prompt. It reads your words as tokens and keeps the conversation’s context in mind. Sometimes it hallucinates, so you check important facts. The newest versions are multimodal and some can act as agents. That’s the whole landscape in one paragraph.
Why this matters
You don’t need to be technical to use AI, but knowing these words means you can follow any conversation, read any article, and stop feeling left out. That confidence is half the battle. Nobody is born knowing this vocabulary. Every expert you’ve seen confidently tossing these words around learned them the same way you just did, one definition at a time. You’re now caught up.
Frequently asked questions
Do I need to memorize all of these?
No. The big four are prompt, model, hallucination, and context. The rest you’ll pick up naturally.
What’s the difference between ChatGPT and an LLM?
ChatGPT is a product. An LLM is the type of AI underneath it. ChatGPT is built on an LLM, the way a car is built on an engine.
Why does AI “hallucinate”?
It predicts likely answers rather than looking up guaranteed facts, so sometimes it predicts something that sounds right but isn’t. Always verify important details.
Is “prompt engineering” a real job?
It’s a real skill, and some people do it professionally. But for everyday use, it just means asking clear, specific questions.
What’s an AI agent in simple terms?
An AI that can do things for you, not just talk. Think of a chatbot that can also act, like actually booking the appointment instead of just telling you how.