GR Stories | How to Write the Perfect AI Prompt (And Actually Get What You Need)
Most students use AI every day. Very few know how to talk to it properly. Here is what changes when you do.
There is a specific kind of frustration that comes from asking an AI a question and getting an answer that is technically correct, completely generic, and entirely useless.
You wanted help summarising a complex chapter in a way that makes sense for your exam. You got five paragraphs that sounded like a Wikipedia article with better punctuation. You asked for feedback on your essay. You got eleven generic suggestions, none of which addressed the actual problem. You asked for a study plan. You got a template that could have been written for anyone, anywhere, in any situation.
It is not that the AI failed. It is that the question you gave it did not have enough to work with.
This is the part most people miss. AI tools like ChatGPT, Claude, or Gemini are not search engines. They are not databases you query with keywords. They are closer to a very well-read person who will do their best with exactly what you give them. If what you give them is vague, rushed, or incomplete, what comes back will reflect that.
The good news is that the fix is not complicated. Writing a better prompt is mostly a matter of understanding what a prompt actually needs.
What a prompt actually is
A prompt is not just a question. It is the full context you give the AI before it starts writing.
Everything you include — or do not include — shapes the response. The role you assign the AI, the format you ask for, the constraints you set, the specific outcome you want: all of these are part of the prompt, whether you state them explicitly or leave them out.
When you leave things out, the AI fills in the gaps by making assumptions. Those assumptions are not always wrong, but they are rarely exactly right for your situation. The more specific you are, the less the AI has to guess.
The four things every good prompt has
You do not need to follow a rigid formula. But most genuinely useful prompts share four qualities.
1. A clear role or frame
Tell the AI what it is supposed to be in this exchange. Not in a theatrical way — just enough to set the right mode.
"Act as a professor explaining this concept to a second-year student" produces something different from "explain this concept". The first has direction. The second leaves the AI guessing the appropriate register, depth, and vocabulary.
For students, this is especially useful. You can tell the AI to think like a tutor, an examiner, a classmate who already understood the material, or a patient editor who will not flatter you. Each framing changes the output meaningfully.
2. Specific context about your situation
This is the part most people skip, and it is the one that makes the biggest difference.
The AI does not know you are writing in your second language. It does not know your exam is in three days, that your professor prefers analytical arguments over descriptive ones, or that you are working from a specific text. Unless you say so.
Context does not mean over-explaining. It means giving the AI the relevant details it cannot know by itself:
- what you are studying, at what level, and where
- what the final output is for (an exam, an email, a presentation)
- any constraints that matter (word count, tone, format)
- what you already have or have already tried
A prompt with good context produces an answer that fits your actual situation. Without it, you get a general answer that might be fine for most people and not quite right for you.
3. A precise request — not just a topic
There is a difference between asking about a topic and asking for a specific thing.
"Tell me about monetary policy" and "Explain how central banks use interest rates to control inflation, in plain language, in about 200 words" are both about monetary policy. But they will produce very different responses. The second one has a purpose, a scope, and a format. The first leaves every one of those decisions to the AI.
Precision is not about being demanding. It is about giving the AI fewer decisions to make on your behalf.
4. The format you want
AI tools default to a format when you do not specify one. Sometimes that default works. Often, it does not match what you actually need.
If you want bullet points, say so. If you want a structured table, ask for a table. If you want a flowing explanation you can read like a short essay, say you want flowing prose, no headers. If you want the response to be short, give a word limit.
The model will not judge you for being specific. It will simply do what you asked.
A before and after
Here is how the difference looks in practice.
Before: "Help me study for my economics exam."
This is a topic with a vague request. The AI does not know what the exam covers, when it is, what level you are studying at, or what kind of help you actually need. It will probably generate a generic study plan or a broad overview of economics.
After: "I have an economics exam in two days on macroeconomic theory, specifically on IS-LM models and the Keynesian cross. I am studying at undergraduate level in Italy. Can you act as a tutor and explain the key concepts I need to understand, one at a time, with a practical example for each? Use plain language. I do not need definitions I can find in a textbook — I need to understand the logic."
That second prompt has a role, context, a specific request, and a format preference. The AI now has something real to work with. The answer will not look like a Wikipedia article.
The thing most people do not do: push back
A prompt is not always a single message. It is often a conversation.
If the first response is not right — too long, too shallow, too formal, misses the point — do not start over. Correct it. Tell the AI what is wrong with the output and what you need instead.
"This is too academic. Simplify the language."
"You covered the theory but skipped the application. Give me examples."
"This is 400 words but I need 150. Cut the fluff."
"You explained what I asked, but you didn't explain why it matters. Add that."
AI tools are genuinely good at course-correcting when you tell them exactly what is off. Most people just accept the first answer. The people who get the most out of these tools are the ones who treat the conversation as iterative — not as a single question with a single answer.
Specific prompts that actually help students
These are not templates. They are starting points you can adapt to your situation.
For understanding difficult material: "I am studying [topic] for a [level] course. I read the following passage but I cannot follow the logic: [paste text]. Explain what it is trying to say in plain language, and tell me the one thing I actually need to remember from it."
For essay feedback: "Read this essay draft and tell me what the weakest parts are. Be direct — I do not need encouragement, I need an honest critique. Focus on structure and argument, not grammar. Here is the draft: [paste text]."
For summarising readings: "Summarise the key argument of this text in three to five bullet points, written for someone who needs to use this for an exam. Do not copy sentences from the text. Reframe the ideas in your own words: [paste text]."
For preparing for an oral exam: "I have an oral exam on [topic]. Simulate the role of my examiner. Ask me one question at a time, then wait for my answer before asking the next. If my answer is incomplete or vague, tell me what I missed. Start with the first question."
For writing emails in a foreign language: "I need to write a formal email in [language] to my university professor. I am writing to request an extension on my essay deadline because [reason]. I am not a native speaker, so the language should be correct but not overly elaborate. Draft the email and explain the key phrases I should know."
For building a study plan: "I have [X] days before my [subject] exam. The main topics are [list]. I study best in short sessions with breaks. I need about [X] hours of revision per day. Can you build me a day-by-day plan that covers all the topics without cramming everything at the end?"
What to avoid
A few habits that produce consistently poor results:
Being too vague. The shorter and more general the prompt, the more the AI will fill in the gaps with guesses. Short is not the same as precise.
Asking multiple unrelated things at once. If you need an essay summary, a study plan, and an email draft, do them separately. Combining them usually means each one is done badly.
Accepting the first answer without reading it carefully. AI responses can sound confident even when they are wrong, off-topic, or not quite what you needed. Read critically, the same way you would read anything.
Using it to think for you instead of with you. There is a real difference between using AI to replace your thinking and using it to accelerate it. One makes your work worse in ways that are hard to see immediately. The other makes you faster and often sharper.
Ignoring the output format. If you need a table, a list, or a short paragraph and you do not say so, you will spend time reformatting an answer that could have arrived in the right shape.
A note on trust and verification
AI tools get things wrong. They can generate plausible-sounding facts that are incorrect, cite sources that do not exist, or explain concepts with enough confidence to make a mistake feel authoritative.
This is not a reason to avoid them. It is a reason to use them with the same critical eye you would apply to any source. If you are working on something that matters — an essay, an exam, a real decision — verify what the AI tells you against something you know is reliable.
AI is a tool for thinking faster and more clearly. It is not a replacement for knowing whether what you are reading is true.
The bigger shift
The students who get the most out of AI are not the ones who found a magical prompt. They are the ones who learned to think about what they actually need before they ask for it.
A good prompt is, in some ways, a precise question. And asking a precise question requires you to have already done some of the thinking. You need to know what you are trying to achieve, what you have already tried, and what specific gap you need to fill.
That process — defining the problem clearly before seeking a solution — is useful far beyond AI. It is a study skill. A writing skill. A thinking skill.
The AI just gives you a reason to practise it every day.
