The Simplest Possible Definition
A prompt is everything you send to an AI model before it responds. That includes your question, any instructions you give it, any examples you provide, and any background context you share. The model reads all of it, then generates a response that continues naturally from what you wrote.
Think of it like giving instructions to a very capable but very literal colleague. If you say "write me an email," you might get a generic template. If you say "write me a two-paragraph email to a client who has not responded in three weeks, using a friendly but firm tone, asking them to confirm the next step by Friday," you will get something you can actually use. The model did not become smarter. You became a better communicator.
Why Does Wording Change the Answer?
A language model generates its response one word (technically one token) at a time. At each step, it looks at everything written so far and predicts the most likely next word. The prediction is shaped entirely by patterns learned from the enormous amount of text the model was trained on.
This means the model does not "understand" your question the way a human does. It reads the pattern of your words and continues it. "Summarize this article" and "give me the three most important takeaways from this article in bullet points" both ask for a summary. But the second one triggers a pattern closer to structured summaries. The model has seen thousands of such patterns and produces a more structured output.
This is why prompt engineering is a real skill. You are not programming. You are not writing code. You are communicating in a way that activates the patterns you want.
The Three Parts of Every Prompt
Every effective prompt has three components, whether you write them explicitly or not.
The role or persona. This tells the model who it should be when it responds. "You are an experienced nurse" versus "you are a marketing consultant" will produce very different answers to the same question about communicating with a difficult person. The role activates different knowledge patterns from training.
The task. This is what you want the model to do. "Summarize," "draft," "analyze," "list," "explain," "compare." The more precise your verb, the more precise the output. "Write" is vague. "Write a three-sentence summary" is a task with a built-in format.
The context. This is any background information the model needs to complete the task well. The name of the client. The audience for the document. The constraint on length. The format you expect. The more relevant context you include, the less the model has to guess.
The Context Window: Your Model's Working Memory
The model can only see what you put in the prompt. It has no memory of previous conversations unless you paste them in. It cannot browse the internet unless it has a tool that lets it do so. It cannot look up today's date. Everything it knows comes from its training data, and everything it knows about your specific situation comes from your prompt.
The context window is the maximum amount of text a model can see at once, measured in tokens. A token is roughly three to four characters of English text. Modern models have context windows ranging from tens of thousands to hundreds of thousands of tokens. For most day-to-day tasks this is more than enough. But for very long documents, you may need to chunk the content and process it in pieces.
This matters because every token in your prompt takes up space that could hold more context. Long, rambling prompts waste that space. Short, precise prompts leave more room for the content you want the model to analyze.
System Prompts versus User Messages
Many AI applications have two distinct input channels. The system prompt is set by the developer or application owner before the user ever types anything. It contains standing instructions: the model's role, rules about what it should and should not say, the format it should use, and any background information relevant to every conversation. The user never sees the system prompt.
The user message is what you type in the chat window. In most consumer AI tools you are working only with the user message channel. In professional or developer contexts, you can often set both.
Understanding this distinction matters when you want to build a consistent AI experience for a team or product. A well-written system prompt means every user in your organization gets responses in the same format and tone, without having to re-explain the context each time.