A beginner's guide to writing good AI prompts

Technology · 6 min read

If an AI chatbot has given you a vague, generic or off-target answer, the cause is usually the request rather than the tool. A prompt is simply the instructions you give, and like instructions to a smart new colleague, they work best when they are clear and complete. Here is a practical framework you can use with any modern AI assistant.

Think of the AI as a capable but new colleague

An AI model knows a great deal in general, but it knows nothing about your situation until you tell it. It cannot read your mind, see your files unless you share them, or guess who your audience is. If you would need to explain background to a human helper before they could do the task, explain it to the AI as well.

The five building blocks of a good prompt

1. Goal: what you want done, in one clear sentence.
2. Context: who it is for, why you need it and any relevant background.
3. Constraints: length, tone, things to include or avoid.
4. Format: bullet points, a table, an email, steps or code.
5. Input or example: the text to work on, or a sample of what good looks like.

You do not need all five every time, but the more of them you cover, the less the model has to guess.

Before and after

Weak prompt	Stronger prompt
Write about healthy eating.	Write a 300-word beginner-friendly blog intro on healthy eating for busy office workers in India. Use a friendly tone, mention easy swaps like replacing fried snacks with roasted chana, and end with a question to the reader.
Fix my email.	Rewrite the email below so it is polite but firm. It is to a vendor who is late with delivery. Keep it under 120 words and include a clear deadline of Friday.
Explain machine learning.	Explain machine learning to a 15-year-old in three short paragraphs, using an example from cricket or a streaming app. Avoid jargon.

Give a role or an audience

Telling the AI who it should act as, or who the answer is for, changes the level and style. "Explain this as a patient maths tutor to a Class 8 student" produces a different answer from "Summarise this for a busy manager in five bullets." Be careful, though: a role does not make the model an actual licensed professional.

Show an example

If you want a specific style, show it. Paste one or two examples of the output you like, then ask for new ones in the same pattern. This is called few-shot prompting, and it is often the quickest way to get a format right. For instance, give two sample product descriptions in your brand voice and ask for a third.

Break big tasks into steps

Complex requests often work better when you split them. Instead of asking for a complete business plan in one go, ask first for an outline, then for each section in turn. You can also ask the AI to think through steps, for example, "First list the assumptions, then calculate, then give the final answer." This helps with maths, planning and analysis, and makes errors easier to catch.

Tip: If you are not sure what to ask for, ask the AI to ask you. Try: "Before you start, ask me any questions you need to produce the best result."

Iterate like a conversation

The first answer is a draft. Respond with specific feedback: "Make the tone more casual," "Shorten to half," "Add a real-world example," "Remove the second paragraph." Vague feedback such as "make it better" gives vague results. If the conversation goes off track, start fresh with an improved prompt rather than piling on corrections.

Common mistakes to avoid

• Being too vague: "Tell me about marketing" gives a textbook overview.
• Cramming many unrelated tasks: split them into separate prompts.
• Not specifying the format or length.
• Trusting facts blindly: AI can state wrong information confidently, so verify numbers, quotes, dates and legal or medical claims.
• Sharing sensitive data: avoid pasting passwords, Aadhaar or bank details, confidential company documents or other people's private information into any tool unless you know how it is handled.

A reusable template

Copy this skeleton and fill it in: "I am [who you are] working on [task]. The audience is [who]. Please [goal]. Use a [tone] tone and keep it to [length]. Format the answer as [format]. Here is the material: [paste]."

Final thoughts

Good prompting is a small skill that improves with practice. Start with a clear goal, add context and constraints, review the output critically and refine. Treat the AI as a helper whose work you check, not a source of unquestionable truth. Different tools behave a little differently, so experiment to learn what works best with the one you use.