Prompt

Latest update: 26/04/27

Definition

A prompt is the input you give an AI model – the text, question, instruction, or context that tells it what to do and shapes the response it generates.

What Is a Prompt?

A prompt is what you type (or send) to an AI. It’s the starting point for everything the model produces. Ask a question, give an instruction, paste a document with a task – all of that is your prompt.

The word comes from theater: a “prompt” is a cue that tells an actor what to say or do next. AI prompts work the same way. You cue the model, and it responds.

Without a prompt, an AI model does nothing. It has no built-in agenda, no spontaneous thoughts. Everything it produces starts with input from you. That’s what makes prompt quality so important – the model is always, in some sense, a reflection of what you gave it to work with.

💡 How Does It Work?

When you submit a prompt, the model converts it into tokens, processes them through its neural network, and generates a response – one token at a time – based on its training and the content of your input.

The model reads the entire prompt before it generates anything. That means the order, structure, and wording of your prompt all affect the output. An analogy: think of a prompt like a job brief you hand to a contractor. A vague brief (“build me something nice”) gives them too much latitude. A specific one – with scope, audience, constraints, and format – gets you closer to what you actually had in mind.

The model is also pattern-sensitive. If your prompt resembles the structure of a question, it tends to respond like an answer. If it resembles the opening of a document, the model often continues the document. What you start shapes what comes next.

Why It Matters for Your Prompts

Every frustrating AI interaction – generic answers, missed instructions, wrong tone, wrong format – traces back to the prompt in some way. The model isn’t being difficult. It’s responding to the signals you gave it. Change the signals, and the output changes.

A prompt isn’t just a question. It’s a complete communication: who the AI should act as, what you need, who it’s for, what format it should follow, and what to avoid. Most people use only a fraction of that space.

Compare these two prompts:

“Write a summary.”
“Write a 3-sentence executive summary of the following meeting notes for a non-technical audience. Focus on decisions made and next steps. Leave out the discussion backstory.”

The model’s capabilities are identical for both. The second prompt does more of the thinking upfront – and that work pays off in the output.

🌐 Real-World Example

A small business owner wants help writing a sales email. Her first attempt: “Write me a sales email.”

The result is generic – “Dear Valued Customer, We’re excited to offer you…” She’d seen it a hundred times and couldn’t use it.

She tries again with more structure: “Write a short sales email (under 150 words) for a specialty coffee subscription service targeting busy professionals. The offer is a 20% discount on the first month. Tone: warm but not salesy. Call to action: click to claim the offer.”

The second prompt took two extra minutes to write. The email it produced took two seconds to approve. The only thing that changed was the quality of the input.

Related Terms

Prompt Engineering – The practice of crafting and refining prompts to consistently get better results.
System Prompt – A special type of prompt set by developers or platform operators that shapes how the AI behaves before the user types anything.
Context Window – The total space available for your prompt plus the AI’s response; longer prompts leave less room for output.
Temperature – A setting that affects how the model responds to your prompt – more predictably or more creatively.

Encyclopedia Prompting Techniques Architecture & Technical Advanced Concepts

Frequently Asked Questions

Is there a “correct” way to write a prompt?

There’s no single format, but there are consistently useful elements: role (who should the AI be?), task (what exactly should it do?), context (what does it need to know?), format (how should the output look?), and constraints (what should it avoid?). You don’t need all five every time, but each one you add tends to improve the result.

Does prompt length matter?

Length matters less than precision. A long, rambling prompt can confuse the model just as much as a short, vague one. The ideal prompt is as long as it needs to be to communicate your intent clearly – and not a word longer. For complex tasks, that might be several paragraphs. For simple ones, two sentences is often enough.

Why does rewording my prompt change the answer so much?

Language models are sensitive to how things are phrased. Different wording activates different patterns from training. Asking “what are the benefits of X?” pulls in positive framing. Asking “what do critics say about X?” pulls in a different set of patterns. Neither is wrong – they’re just asking different things. This sensitivity is exactly why prompting is a learnable skill.

Can I re-use the same prompt for different tasks?

Yes – this is the idea behind prompt templates. A well-designed prompt structure can be adapted across many use cases by swapping out the specific details. If you find a prompt format that works well for a particular type of task, it’s worth saving and reusing it.

References

OpenAI – “Prompt Engineering Guide” – Official guidance from OpenAI on writing effective prompts.
Anthropic – “Prompt Engineering Overview” – Covers Claude-specific prompting strategies and best practices.