How to ask AI for better results

What is a prompt?

A prompt is the message you send the AI describing what you want. The clearer and more specific it is, the better the result — much like briefing a helpful colleague who cannot read your mind. You do not need special syntax or keywords; plain language works well.

A simple recipe for good prompts

Most strong prompts include four things. You will not always need all four, but they are a reliable checklist:

• Context — the background the AI needs: who it is for, what it is about, any constraints.
• Task — the specific thing you want done, stated as an action.
• Format — how you want the answer: length, structure, tone, language.
• Examples — an optional sample of the style or output you have in mind.

Context: I run a small bakery and I'm writing to customers on our mailing list.
Task: Write a short announcement that we now open on Sundays.
Format: Warm and friendly, under 100 words, with a clear closing line.
Example tone: "We've got good news to share..."

Weak versus strong prompts

The same goal produces very different results depending on how you ask:

Write about our new opening hours.

Write a friendly 80-word email to our customers announcing that our bakery is now open on Sundays from 9am to 2pm. End with a warm invitation to visit.

The strong version tells the AI the audience, the key facts, the length, the tone and how to end. There is far less room for it to guess.

Refining answers step by step

You rarely need the perfect prompt on the first try. Treat the first answer as a draft and steer it with short follow-ups in the same conversation:

• “Make it shorter and more formal.”
• “Keep the second paragraph, rewrite the rest.”
• “Give me three alternative subject lines.”
• “Explain the third point in simpler terms.”

Pitfalls to avoid

• Being vague and hoping the AI guesses what you meant.
• Asking several unrelated things in one prompt — split them up.
• Leaving out the audience or the purpose.
• Not saying how long or in what format you want the answer.
• Assuming the answer is correct without checking the facts.