A prompt is the text you give an AI system to tell it what you want. It can be a question, an instruction, a block of data to analyse, or a combination. The quality of the answer depends heavily on the quality of the prompt, which is why writing clear prompts has become a practical everyday skill.
Why prompts matter
AI language models do not read minds. They continue and respond to the text they receive. If your request is vague, the model has to guess your goal, audience, length and format, and its guess may not match yours.
Compare:
- Vague: "Write about solar energy."
- Clear: "Write a 150-word explanation of how rooftop solar panels work for a homeowner in Nepal with no technical background. Use simple words and finish with one tip for choosing an installer."
The second prompt gives topic, length, audience, tone and a closing requirement, so the answer is far more likely to be useful on the first try.
The five parts of a good prompt
You do not need all five every time, but this checklist covers most situations.
- Task: what exactly should the AI do? Summarise, explain, rewrite, translate, brainstorm, compare.
- Context: background the AI needs, such as who the text is for or what it is about.
- Input: the material to work on, pasted clearly, perhaps between quotation marks or labelled sections.
- Constraints: length, tone, reading level, things to avoid.
- Output format: bullet list, table, email, JSON, step-by-step plan.
A simple template:
Task: Summarise the text below.
Audience: Busy shop owners.
Constraints: Maximum 5 bullet points, simple English, no jargon.
Format: Bullet list, then one sentence recommending a next step.
Text:
"""
...paste your text here...
"""Before-and-after examples
Example 1: emailing a supplier
Weak: "Write an email to my supplier."
Better: "Write a polite, short email to a supplier asking for a revised quote for 200 units, because the last price was 10% higher than agreed. Ask for a reply by Friday. Keep it under 120 words."
Example 2: studying
Weak: "Explain photosynthesis."
Better: "Explain photosynthesis to a Class 8 student in 6 simple steps, with one everyday analogy, then give 3 quiz questions with answers."
Example 3: working with data
Weak: "Analyse this."
Better: "Below are monthly sales figures. Identify the best and worst months, describe any trend in two sentences, and suggest two questions I should investigate. Present results in a small table followed by the two sentences."
Useful techniques
- Give examples. Showing one or two samples of the output you want (few-shot prompting) greatly improves consistency.
- Ask for steps. For logic or maths, asking the model to work step by step often improves accuracy.
- Set a role. "Act as a careful proofreader" can steer tone and focus, though it does not give the AI real expertise.
- Iterate. Treat the first answer as a draft. Follow up with "make it shorter" or "use a friendlier tone".
- Ask it to ask you. "Ask me up to three questions you need before answering" helps when you are unsure what details matter.
Common mistakes
- Being too short and leaving key details out
- Asking several unrelated things in one message
- Not stating the audience or format
- Trusting factual claims without checking
- Pasting private or confidential data into tools you do not control
A quick checklist before you press send
- Is the task a single, clear verb?
- Did I say who the answer is for?
- Did I paste the actual text or data to work on?
- Did I state length, tone and format?
- Is there anything the AI must not do, such as inventing facts or using jargon?
If you can answer yes to these, your prompt is already better than most.
Prompts in apps and code
Developers use prompts inside software too. A typical call to a language model API passes a system instruction and a user message. Here is a simple pattern in Python:
import anthropic
client = anthropic.Anthropic() # reads ANTHROPIC_API_KEY from the environment
response = client.messages.create(
model="claude-sonnet-5-5",
max_tokens=300,
system="You are a concise assistant. Answer in plain English for beginners.",
messages=[{"role": "user", "content": "Explain what an API is in three sentences."}],
)
print(response.content[0].text)The same principles apply: clear instructions, defined audience and a limit on the output.
Frequently Asked Questions
Do longer prompts always work better?
No. They should be as long as needed to be clear, not longer. Remove details that do not affect the result.
Does politeness matter to an AI?
Not for correctness. Clarity matters more than courtesy, although a respectful tone is a good habit.
Should I write prompts in English?
Many models work best in English but also handle Nepali and other languages. Test both for your task and compare the results.
Can a prompt guarantee a correct answer?
No. Prompts improve the odds, but you should still verify important facts, numbers and citations.
Conclusion
A prompt is simply a clear request. State the task, give context and input, set constraints and choose the output format, then refine through follow-ups. That small habit turns a generic chatbot into a much more useful assistant.