Walk into a store and say only, “Recommend a bag.”
The assistant can answer, but she has to guess.
Now try: “I need it for work, it must fit a 13-inch laptop, I do not want a heavy shoulder bag or a loud logo, and my budget is under three thousand dollars.”
Suddenly she has many more useful clues.
That is the simplest way to understand a prompt.
A prompt is the information you give the AI
For everyday use, a prompt can include your question, goal, background, constraints and desired format.
“Write an email” gives very little information.
“Write a polite email under 150 words to my professor asking whether Tuesday works for a thesis meeting; keep it warm but not overly formal” gives much more context.
It is not better because it is longer. It is better because the useful information is there.
Why does the prompt change the answer?
A Large Language Model (LLM) generates text based on what has already appeared. The information that comes before the answer forms the context that influences what the model considers likely or appropriate next.
A store assistant who only knows “I want a bag” may show you the bestseller. If she knows you commute by train, have shoulder pain and dislike large logos, she can eliminate many popular-but-wrong choices.
A good prompt is not necessarily a long prompt
Before writing a giant paragraph, ask four simple questions:
- What do I want the AI to do?
- What background does it need?
- What limits matter?
- What should the answer look like?
Useful constraints often matter more than fancy wording.
Do you need to memorize “perfect prompt” templates?
No.
Templates can remind you to include roles, formats or evaluation criteria, but the most important idea is simpler: the model cannot know requirements you never told it.
If you forgot to mention your budget, the assistant did not technically break your request by showing you a fifteen-thousand-dollar bag.
A prompt is not a guarantee
Even an excellent prompt cannot guarantee factual accuracy.
A prompt helps the model understand your intention. It does not turn every answer into verified truth.
That brings us to Lesson 006: hallucination, when generative AI can produce something wrong while sounding completely confident.
One thing to remember today
A prompt is the task, background and constraints you give an AI. The clearer the useful information, the better the chance that the answer matches what you actually need.