AI
How to Write Good AI Prompts: A Beginner's Guide With 15 Examples You Can Copy
By Abhiram Chandramohan · · 4 min read
Short answer: a good AI prompt gives four things: context (who you are and the situation), the task (exactly what you want), the audience (who it is for), and the format (what the answer should look like). Then you improve the result by replying with feedback, the way you would with a colleague. That covers most of what beginners need to get consistently useful answers from ChatGPT, Claude or Gemini.
Why do beginners get average answers?
Because they type the way they search Google: "marketing plan bakery". The AI has to guess everything else, so it produces something generic. AI assistants are closer to a capable new colleague than a search engine. The more you brief them, the better they perform.
A simple framework: C-T-A-F
- Context: "I run a small home bakery in Bangalore. Most orders come through Instagram."
- Task: "Create a simple marketing plan for Diwali."
- Audience: "Customers are mostly working families in South Bangalore."
- Format: "A week-by-week table with three actions per week. Keep the budget under ₹5,000."
Compare that with "marketing plan bakery". The difference in output is dramatic.
Five habits that improve every prompt
- Give examples. "Write it in this style:" followed by something you like.
- Ask for options. "Give me three different versions."
- Set constraints. Word limits, reading level, tone, what to avoid.
- Ask it to ask you. "Before you answer, ask me any questions you need."
- Iterate. "Shorter." "More formal." "Make point 2 more specific." Refining is faster than restarting.
Common mistakes
- Being vague about the goal.
- Pasting a huge document with no instructions about what to do with it.
- Trusting facts and figures without checking.
- Giving up after one weak answer instead of giving feedback.
- Sharing private or confidential information.
15 prompts you can copy
Work
- "Rewrite this email to sound polite but firm. Keep it under 120 words: [paste]."
- "Summarise these meeting notes into decisions, action items with owners, and open questions: [paste]."
- "Turn these bullet points into a one-page report for my manager: [paste]."
- "I have an interview for [role] at [type of company]. Ask me 10 likely questions one at a time and give feedback on my answers."
- "Explain this spreadsheet data in plain English and list three things that stand out: [paste]."
Study and learning
- "Explain [topic] as if I am a smart 15-year-old, then give me five questions to test my understanding."
- "Create a 4-week study plan for [exam], one hour a day, with weekly revision."
- "Here is my essay. Point out the three biggest weaknesses and how to fix them. Do not rewrite it."
- "Give me the key arguments for and against [issue], with the strongest version of each."
- "Translate this into simple Kannada and keep technical terms in English: [paste]."
Everyday life
- "Plan a 3-day trip to Coorg for a family of four on a moderate budget. Include travel time from Bangalore."
- "Plan a week of vegetarian dinners using these ingredients: [list]. Each under 30 minutes."
- "Help me write a respectful message to my building association about [issue]."
- "I want to start exercising at home with no equipment. Create a 4-week beginner plan."
- "Compare these two phones for my needs (camera, battery, budget ₹25,000): [paste specs]."
Practise with feedback
Prompting is a skill that improves fastest when someone shows you what better looks like on your own tasks. Our AI training for beginners in Bangalore is hands-on from the first class, with prompts and workflows you will use every week. New to the whole idea? Start with how to learn AI from scratch.
Frequently asked questions
Do these prompts work in ChatGPT, Claude and Gemini?
Yes. The principles are the same across assistants.
Should I write prompts in English?
English usually gives the best results, but modern assistants understand Indian languages well enough for many tasks.
What is prompt engineering?
It is the practice of writing instructions that get reliable results from AI. For most people, the framework above is all the "engineering" needed.