The basics
What is a prompt?
If you've ever used ChatGPT, Claude, Gemini, or any AI tool, you've already written a prompt. You probably just didn't call it that — and you almost certainly didn't write it in a way that got the best result.
The simple definition
A prompt is the message you send to an AI before it responds. It's your side of the conversation — your instruction, your question, your request. Everything the AI reads before generating its answer.
When you open ChatGPT and type write me a cover letter, that's a prompt. When you ask Claude summarise this article in 3 bullet points, that's a prompt. Even a single word typed into an AI is technically a prompt.
The key thing to understand
Search engines were built to guess what you mean from a few keywords. AI models respond to exactly what you write — not what you meant, not what you implied. The quality of the output is almost entirely determined by the quality of the prompt.
Why most prompts fail
Most people write AI prompts the same way they write Google searches — a short fragment, a vague question, a few words. That habit made sense for search, but it doesn't work for generative AI.
When your prompt is vague, the AI fills in the gaps on its own. It makes assumptions about who you are, what you need, what level of detail is appropriate, and how to format the response. Those assumptions are almost always wrong for your specific situation.
What the AI has to guess when your prompt is vague
A well-written prompt removes all of that guesswork. The AI doesn't have to assume anything — it just executes precisely on what you gave it.
What a good prompt looks like
The same request — written two ways. The difference in output is not marginal.
Vague
write me a blog post
Specific
Write a 600-word blog post introduction for a B2B SaaS audience about why most companies waste money on tools they never fully adopt. Tone: direct and slightly provocative. End with a question that makes the reader want to keep reading.
Vague
explain machine learning
Specific
Explain how machine learning models are trained, as if I'm a business analyst with no technical background. Use one real-world analogy and keep it under 200 words.
Vague
fix my code
Specific
I have a React useEffect that's triggering an infinite loop. The dependency array includes a function that's re-created on every render. Explain why this happens and show me two ways to fix it — one using useCallback and one restructuring the component.
Common questions
Is a prompt the same as a question?
Not exactly. A question is one type of prompt. But a prompt can also be an instruction, a request, a scenario to roleplay, or a document to summarise. The format depends on what you need.
Do longer prompts always work better?
No. A long prompt full of vague or irrelevant information is worse than a short, precise one. Length isn't the goal — specificity is. A well-structured 3-sentence prompt will consistently beat a rambling paragraph.
What's the difference between a prompt and a query?
A query is what you type into a search engine — usually just keywords. A prompt is a full instruction to a generative AI. The AI doesn't rank existing results — it generates a new response based entirely on what you wrote.
Does the AI remember previous prompts?
Within a single conversation, yes — most AI tools keep a memory of the chat. Across different sessions, typically no. Each new conversation starts fresh unless the tool has a memory feature enabled.
Now learn how to write one properly.
The 5 elements that turn any vague prompt into one the AI can actually work with. Free guide, takes 5 minutes.