Prompt: the instruction you give an AI
A prompt is the text you write to an AI to get a response: a question, an instruction, a document to work on. The clearer and more complete it is, the more likely the result matches what you had in mind.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|
A simple definition
The prompt is everything you give the model to work with: your question, but also any pasted documents, examples, and notes on tone or format. It takes up space in the context window, the amount of text the model can consider at once.
Behind the scenes, assistants often add their own standing instructions, called a system prompt, which set the tool’s overall behavior. You can sometimes add your own preferences through custom instructions.
Before and after
Vague prompt: “Write an email to my client.”
Specific prompt: “I run a small plumbing business. Write a short, friendly email to a client letting them know that Thursday’s appointment has to move to Monday because a part is back-ordered. Offer two time slots and end with an apology. Professional tone, 120 words max.”
The second prompt gives the context (who you are), the goal (announce a delay), the constraints (time slots, length) and the tone. The model no longer has to guess.
Why it matters to you
An AI model knows nothing about your job, your client or your habits. OpenAI’s official guide sums it up: prompt engineering is the process of writing effective instructions so that a model consistently generates content that meets your requirements (OpenAI guide). Spending a few extra seconds on the request often saves several rounds of back-and-forth. The prompt generator walks you through structuring a request step by step.
Common mistakes
- Leaving out context: the model fills the gaps with assumptions.
- Asking for too much at once: splitting the job into steps works better.
- Not specifying the format: list, table, length, language.
- Pasting sensitive data without checking the tool’s privacy settings.
- Trusting outside text blindly: a web page or document can hide instructions that hijack the AI. This is called prompt injection.
To go further, prompt engineering covers more advanced techniques such as examples, step-by-step reasoning and structured output formats.
Related terms
- Prompt engineering · The practice of writing, testing and refining instructions to an AI model so its answers become more reliable and useful.
- System prompt · Background instructions given to a model before the conversation starts (role, tone, rules) that shape every answer that follows.
- Custom instructions · Preferences you save once in an AI assistant (who you are, tone, format) that it then applies to every conversation.
- Few-shot prompting · A technique where you include a few worked examples of input and expected output in the prompt to show the model what you want.
- Context window · The maximum amount of text (in tokens) a model can read and produce at once, including its own answer.
- Prompt injection · An attack that slips hidden instructions into text an AI reads, to make it ignore its original instructions or act against its user.
Try it
Full glossary
Frequently asked questions
Is there a right way to write a prompt?
There is no single formula, but good prompts share the same ingredients: the relevant context, a clear goal, any constraints and the format you expect. If the answer misses the mark, add the missing detail rather than starting from scratch.
Should prompts be long?
Not necessarily. A prompt should be complete, not long. Three well-built lines often beat a confusing page of text.
Are there magic phrases that unlock better answers?
No. Tricks like « you are the world's best expert » have a limited and inconsistent effect depending on the model. What works best is describing the situation clearly, giving an example and saying exactly what you want back.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.