미래가치를 창출하는 글로컬 산학일체 혁신대학
Same AI, Different Results:
How Prompts Make the Difference
By Seo Jin-Jeong, Reporter, Senior of Computer Engineering
These days, AI has become a familiar tool that many students use at least once while working on assignments. When organizing a presentation script, planning the flow of a report, or finding the cause of an error in a coding task, many people turn to AI first. But even when people use the same AI, the results are not the same for everyone. Some people say, “This is good enough to submit right away,” while others feel, “This is not the answer I wanted.” AI may seem capable of handling a wide range of academic tasks, but in reality, the results depend more on how the user asks than on the AI itself.
The key thing that creates this difference is the “prompt.” A prompt is the question or request you type into AI. In other words, it is the set of words you use to tell AI what you want and under what conditions. Even with the same topic, if the prompt is specific, the answer is organized to fit the user’s goal. If there is not enough information, the answer may go in a different direction from what the user intended. That is why the result changes depending on the prompt, even when people use the same AI. In this article, we will find out why results change depending on how prompts are used and go over practical prompt-writing tips with examples.
Six Prompt Rules That Improve the Quality of Answers
① Set the role first
If you first tell AI from what perspective it should respond, the answer becomes much more realistic. This is called a “persona.” A persona is the character and point of view of the person answering. Even with the same question, the standard and wording can change depending on who is speaking. If you do not set a role, AI tends to give textbook-style advice, and the response often ends up being correct but not very helpful. On the other hand, if you set a role, AI builds its answer by following the standards and language of that role.
Bad prompt: “What mistakes are in my resume?”
Good prompt: “You are a hiring manager at a large company. Based on how you review resumes from new developers, please tell me five mistakes applicants often make.”
When you set the role like this, AI gives answers based on the evaluator’s check points instead of speaking like a supportive friend.
② Explain who needs the answer
If you tell AI who the answer is for, it adjusts the level and wording. This works especially well when the situation is clear, such as for students or job seekers. The reason to name the audience is that even with the same topic, beginners and experienced people need different kinds of information. If there is no information about the audience, AI may keep expanding the explanation, and the main point can become unclear. But when the audience is clear, AI can organize the response by prioritizing the advice the reader needs most at that moment.
Bad prompt: “How do I write a resume?”
Good prompt: “I am a job seeker writing a resume for the first time. I am preparing to apply for a developer position.”
If you ask like this, AI will try to match the answer to a new applicant instead of using overly advanced terms.
③ Show the tone you want with an example
If there is a writing style you want, giving just one or two sample lines can make the result much more organized. AI is good at following examples. The reason examples help is that they automatically guide the style and tone. This is especially important in Korean writing, where an awkward tone can stand out easily. If the user shows the kind of sentence they want in advance, AI can avoid unnecessary exaggeration or stiff expressions and make the writing flow more naturally.
Bad prompt: “Please make the sentences simple.”
Good prompt: “Please write in short, clear sentences with fewer than 20 characters each. Please follow the examples below.
ex1) I was in charge of teamwork in the project.
ex2) I solved the problem to the end.”
④ Give your own situation as material
AI does not know the user’s background. So for writing that needs personal detail, like resumes or self-introduction letters, background information is essential. If you do not give this information, AI will combine common phrases, and the result can sound generic. But if you give your real background, AI can connect those details and create writing that sounds more like “your own story.”
Bad prompt: “Please write my reason for applying.”
Good prompt: “I graduated from Hanbat National University with a degree in Computer Engineering. I have won awards in competitions. I worked on the front-end in a team project. Based on this, please write a 7-sentence paragraph about my reason for applying.”
When you give material like this, AI does not fill in the blanks with guesses. Instead, it builds logic based on the experience you provided.
⑤ Tell AI to organize the answer in order
Instead of asking for everything at once, providing instructions in a clear order makes the answer much easier to follow. This is especially helpful in job preparation, because there are many things to do, so it is better to write prompts in steps. Step-by-step prompts work well because they help AI present information in a clear sequence. A simple explanation may still leave the user confused about what to do first. But when steps are included, the user can follow the response more easily like a checklist. As a result, it takes less time to ask questions, and it becomes easier to turn the answer into real action.
Bad prompt: “What should I do first to write a resume?”
Good prompt: “I am a job seeker writing a resume for the first time for a new developer position. Please explain the resume-writing process in five steps. For each step, also write what I should include and what I should be careful about.”
Bad prompt: “People say my project description is not good. How do I fix it?”
Good prompt: “I often get feedback that my project description in my resume is too abstract. Please first point out the cause, and then suggest how to fix it with example sentences.”
When you ask like this, AI gives a structured answer such as “problem check → way to improve → example to apply.”
⑥ Improve it several times
It is okay if you are not satisfied with the first answer. A prompt is not something you complete in one try. It is a process of refining your request over time. The reason for asking again is that it helps AI reflect the user’s preference more accurately. Writing is often not finished in one draft, and job documents especially need to be improved over time by matching “my situation” with “the evaluation standard.” Through repeated questions, you can make the sentences shorter, more specific, or better in tone, and improve the final result.
Bad prompt: “I don’t like it. Write it again.”
Good prompt: “Please make it a little shorter.”
“It sounds too stiff. Please make it sound more natural.”
“Please change this sentence so it is easier to explain in an interview.”
AI is not a tool that finds the right answer on its own. It is a tool that creates results based on the conditions given by the user. So the more specific the prompt is, the more consistent the result becomes. The key process is to set a role and give a point of view, name the audience and match the level, provide examples and background information to make the writing realistic, create a step-by-step flow so it can be used in real action, and then improve it through repeated questions when needed. If you learn just this process, you can get very different results from the same AI. A person who uses AI well is not someone who simply gets good answers, but someone who designs good questions.
https://www.elancer.co.kr/blog/detail/250
https://brunch.co.kr/@5dcc225369644b6/39
https://www.youtube.com/watch?v=ruJ3oxXcNJQ