A prompt is the input you give an AI model: the words you type into the chat box that tell it what to do. It solves a specific problem. It turns the rough idea in your head into a concrete request the model can act on, and the clearer you write it, the closer the result lands to what you wanted.
The one-sentence definition
A prompt is the instruction you write for an AI, and the model reads it to figure out what to generate, in what format, and for whom. It can be as short as “translate this” or as long as a full page that packs in background, examples, constraints, and the output format you want. A command, a question, the reference material you paste in, and the tone you’re after: any of these can be part of a single prompt.
A concrete example
Here is an example you can copy directly: in ChatGPT, you type “turn the meeting notes below into 5 to-dos, each under 20 words, sorted by priority,” and that whole sentence is a prompt. It carries four parts: the task (make to-dos), the quantity (five), the constraint (under 20 words each), and the sort order (by priority). Type “help me sort out these meeting notes” instead and it still runs, but the output drifts much more. Keep in mind that even a detailed prompt does not guarantee correct answers; the citations and numbers it returns can be wrong, so review important claims against the source yourself.
How a prompt differs from context and system instructions
A prompt is what you actually send this turn, context is what the model still remembers from earlier turns, and a system instruction is the underlying rule the developer set in advance that you usually never see. The three stack together to shape the result: the same prompt placed after different context can produce very different answers. In Claude, drop a long document in and then ask a question; the document becomes context, and your question is the prompt. One caution here: conclusions it draws from web findings or the material you supplied still need a check against the original source before you treat them as settled.
When you actually use one
Any time you type into an AI chat box, you are writing a prompt; the only difference is whether you write it casually or with care. A quick everyday question needs nothing fancy. When you want output you can use as is — a well-worded email, a block of code with a fixed structure, a report filled into a template — it pays to state the task, format, examples, and limits in one go. To save effort later, save your frequent prompts as templates and swap a few words the next time.
