Practical Notebook for Asking AI the Right Way
A practical notebook that organizes how to convey context, purpose, and output format together in frequently used AI application situations like code modification, writing, summarization, translation, and image analysis.
Asking AI Questions Well
Asking AI a question seems easy.
Write a question, wait for an answer. But when you actually try to get the answer you want, it often goes differently than expected. An answer comes, but it's in the wrong direction, wrong format, or doesn't fit your situation.
The problem isn't always that AI is insufficient.
Often, I haven't given enough information about what I want, what context I need it in, or what form I should receive it in.
This article is less about deeply exploring one concept and more like a practical notebook—a collection of prompt formats you can pull out and use immediately in situations you encounter often.
1. Code Fixes / Debugging
Vague Request
Why doesn't this work? Fix it.
Without code, analysis is impossible. Without an error message, the cause must be guessed. AI imagines code and responds, and if that imagined code differs from reality, the answer won't be correct.
Request with Criteria
This Python function has a bug where it returns None under certain conditions.
[Paste code]
Error message:
[Paste error message]
Situation where this bug occurs:
- When input is an empty list
What I want:
1. Explanation of the bug cause
2. Fixed code (show before/after together)
3. How to avoid this mistake
Code + error message + occurrence situation + desired output format. With these four things, the answer's direction narrows.
Ready-to-Use Template
A problem occurs in the following {{language/framework}} code.
[Code]
{{Paste code}}
[Error Message]
{{Paste error message}}
[Occurrence Situation]
{{What conditions trigger the error}}
[Environment]
- Language version: {{e.g., Python 3.11}}
- Major libraries: {{e.g., FastAPI 0.100}}
What I want:
1. Root cause analysis
2. Fixed code (with comments explaining changes)
3. Points to watch out for in this pattern
2. Code Review
Vague Request
Review this code for me.
It doesn't know what perspective to use for the review. Is it performance? Readability? Security? AI will likely give general feedback without direction.
Request with Criteria
Review the following Python FastAPI code.
[Code]
{{code}}
Review perspectives:
1. Security vulnerabilities (SQL injection, missing authentication, etc.)
2. Is error handling appropriate?
3. Parts that can be performance-optimized
Format:
- For each item: "Issue / Risk Level (High/Medium/Low) / Fix Suggestion"
- Include corrected code examples for things that need fixing
- Don't mention parts that are already well done
Ready-to-Use Template
Review the following {{language/framework}} code.
[Code]
{{code}}
[Review Perspectives] (keep only what applies)
- Security vulnerabilities
- Performance optimization
- Readability / code quality
- Error handling
- Test coverage
[Format]
For each issue: Issue / Severity (High/Medium/Low) / Fixed code example
Also add a one-line overall evaluation at the end.
3. Writing (Blog, Articles)
Vague Request
Write a blog post about AI.
Who is it for? What tone? How long should it be? Without anything, you get a generic-sounding post like something you've seen before.
Request with Criteria
Write a blog post under these conditions.
Topic: How to Make a Chatbot with Claude API
Target readers: Developers who know Python basics but are trying an AI API for the first time
Purpose: Read this post and build a working chatbot in 30 minutes
Length: Around 1500 characters
Tone: Friendly and practical. Minimize theory, focus on example code.
Structure:
1. Completed screen (described in text)
2. What you need (including pip install)
3. Code explanations step by step
4. Execution result examples
5. Common error solutions
SEO keywords: Claude API, Python chatbot, AI API integration
Ready-to-Use Template
Write a blog post under these conditions.
Topic: {{topic}}
Target readers: {{e.g., non-developer office workers / beginner developers / senior developers}}
Purpose: {{what readers should gain from this post}}
Length: {{character count or word count}}
Tone: {{e.g., professional / friendly / humorous}}
Structure:
- Introduction: {{how to start}}
- Body sections: {{what content should be included}}
- Conclusion: {{how to end}}
Include: {{e.g., code examples / tables / real cases}}
Avoid: {{e.g., overly technical terms / theory-focused}}
SEO keywords: {{3 main keywords}}
4. SNS / Social Media Content
Vague Request
Write an Instagram post.
What kind of account? What content? How many hashtags? Without anything, you get a post that could work for any account.
Request with Criteria
Write a post for my Instagram account (@_dev, AI/development content).
This post's content:
"Good prompts make work 10x faster"
Tone: Practical and relatable. Not too salesy.
Length: Caption around 150 characters
Hashtags: 10 relevant hashtags
Include a closing question to encourage follower engagement
Emojis: Use naturally
Ready-to-Use Template
Write a {{platform}} post.
Account character: {{e.g., AI/development info account / lifestyle / food blog}}
Post topic: {{topic or message to convey}}
Tone: {{e.g., friendly / professional / humorous / motivational}}
Length: {{character/word count or "short/medium/long"}}
Include:
- {{number}} hashtags
- {{emoji usage yes/no}}
- CTA: {{likes / comments / link clicks}}
5. YouTube / Shorts Script
Vague Request
Write a YouTube Shorts script.
What topic? How many seconds? Who's it for? Without anything, you get a plain script. In Shorts, if you don't hook viewers in the first 3 seconds, they scroll down. This 3-second hook is the most critical, but "write a script" alone can't create that properly.
Request with Criteria
Write a YouTube Shorts script.
Channel: AI Usage Tips Channel (50K subscribers, target 20-40 year old office workers)
Topic: "Ask ChatGPT this way and you'll get 3x better responses"
Length: 55 seconds (about 165 words at speaking pace)
Tone: Fast and practical. Friendly casual speech.
Structure:
- Hook (0-3 seconds): Shocking fact or question that makes viewers stop
- Problem (3-15 seconds): How most people misuse AI
- Core content (15-45 seconds): 2-3 specific tips with examples
- CTA (45-55 seconds): Encourage subscriptions or comments
Additional conditions:
- Include key phrases with [subtitle] marking
- Include simple visual direction cues with (screen)
Ready-to-Use Template
Write a {{video type: Shorts/regular}} script for YouTube.
Channel character: {{channel topic and main audience}}
Video topic: {{topic}}
Length: {{seconds or minutes}}
Tone: {{e.g., friendly / professional / humorous}}
Speech style: {{e.g., formal / casual / mixed}}
Structure:
- Hook ({{seconds}}): {{how to start}}
- Main content {{seconds}}: {{core message}}
- CTA {{seconds}}: {{desired action: subscribe / comment / like}}
Include:
- [Subtitle] marking for emphasis keywords
- (Screen) marking for visual directions
6. Email / Business Documents
Vague Request
Write an email. About scheduling a meeting.
Who are you sending it to? What's your relationship? What meeting? Without this, AI writes the most generic email format. Your relationship with the recipient and company culture aren't reflected at all.
Request with Criteria
Write a business email.
Recipient: CTO at a startup I'm contacting for the first time (following on LinkedIn)
Purpose: Request a 30-minute meeting to explore collaboration possibilities
About me: AI consulting company CEO, 3 years experience
Proposed meeting dates: May 8th (Thursday) 2-5 PM, or May 9th (Friday) morning
Format: English, professional but not stiff
Length: Under 150 words, concise
Must include: One sentence on why you want the meeting, one sentence on your value proposition, clear CTA
Ready-to-Use Template
Write in {{email / Slack message / formal document}} format.
Recipient: {{recipient title, company, relationship}}
Sender info: {{your title, company, relevant background}}
Purpose: {{what you want to achieve with this message}}
Tone: {{e.g., formal / friendly / collaborative}}
Length: {{character/word count or short/medium/long}}
Must include:
- {{item 1}}
- {{item 2}}
Avoid: {{e.g., overly long greetings / unnecessary explanations}}
7. Summarization
Vague Request
Summarize this. + [long text]
What's the purpose? One-line summary or item-by-item? Which parts should be highlighted? Without direction, the result might be too long or emphasize unwanted sections.
Request with Criteria
Summarize the following meeting notes.
[Text]
{{meeting notes content}}
Summary purpose: To share with team members who couldn't attend
Format:
- One-line summary (compressed overview)
- Key decisions (bullet points, 3 max)
- Next action items (assignee - task - deadline format)
- Next meeting date (if applicable)
Exclude: Small talk, repeated content, unresolved discussions
Ready-to-Use Template
Summarize the following {{text type: article/meeting notes/paper/report}}.
[Text]
{{content}}
Summary purpose: {{e.g., for presentation / team sharing / personal study / SNS sharing}}
Format:
- {{e.g., one-liner / 5 bullet points / section-by-section}}
- {{e.g., table format / narrative format}}
Length: {{character count or "as concise as possible"}}
Must include: {{key figures / conclusion / action steps etc.}}
Exclude: {{repeated content / examples / footnotes etc.}}
8. Translation
Vague Request
Translate this to English.
What context is this text in? Official document or SNS post? Without this info, you get awkward, word-for-word English.
Request with Criteria
Translate the following Korean text to English.
[Original]
{{text}}
Translation purpose: Business plan for US investors
Tone: Professional and persuasive
Level: Business English, naturally readable by native speakers
Important: No word-for-word translation, localize to English expression style
After translation: If any expressions feel awkward, provide alternatives too
Ready-to-Use Template
Translate the following {{source language}} text to {{target language}}.
[Original]
{{text}}
Translation purpose: {{e.g., email / SNS / official document / presentation}}
Recipient: {{e.g., native speaker / non-native / expert}}
Tone: {{e.g., formal / casual / professional}}
Allow: Localization {{yes/no}}
Additional: {{e.g., provide alternatives for awkward expressions / show original and translation side by side}}
9. Image Analysis
Vague Request
[Upload image] Look at this.
What should I look at? Read text in the image? Evaluate design? Find an error? Without direction, AI doesn't know which elements to focus on when answering.
Request with Criteria
[Upload image]
This is a screenshot of my website's main page.
Analyze from these perspectives:
1. First impression — Can someone see in 3 seconds what this site is about?
2. Readability — Are font size, color contrast, and spacing appropriate?
3. CTA — Is the most important button noticeable?
4. Mobile — Would this look good on mobile? (considering screen size)
For each item: "Current state / Problem / Improvement suggestion" format
Ready-to-Use Template
[Attach image]
This image is {{image description: screenshot/photo/diagram}}.
Analysis perspectives:
{{what you want to analyze}}
Output format: {{by item / table / free form}}
Focus specifically on: {{e.g., top right / text area / graph}}
If anything is unclear, mark it as "unclear" rather than guessing.
10. Brainstorming / Ideation
Vague Request
Give me ideas.
What field? What constraints exist? This basic request yields the most common ideas you can find on the internet.
Request with Criteria
Give me 20 YouTube channel content ideas.
Channel topic: Practical AI tool usage
Main audience: 30-40 year old office workers who want to use AI at work but don't know where to start
Current subscribers: 3,000 (growth stage)
Content that performed well: "Finish report draft with ChatGPT in 30 minutes" (120K views)
Content that didn't perform: AI concept explanation videos (average 500 views)
Conditions:
- Include numbers in titles (e.g., "5 ways", "in 10 minutes")
- Practical, immediately usable content
- Exclude ChatGPT basics that are already widely covered
- Half should be makeable as Shorts
Ready-to-Use Template
Give me {{number}} ideas for {{topic}}.
Context:
- Purpose: {{why you need this idea}}
- Target: {{who it's for}}
- Constraints: {{budget, time, skill level, team size, etc.}}
Reference:
- What worked: {{what was effective in the past}}
- Avoid: {{already tried or unsuitable directions}}
Format:
- Idea title + one-line description
- Difficulty level (Easy/Medium/Hard) marking
11. Data Analysis / Excel
Vague Request
Analyze this data. + [paste table]
What insight do you want? What decision are you trying to make? Without direction, you get a generic statistical summary.
Request with Criteria
Analyze the following monthly sales data.
[Data]
{{paste table}}
Analysis purpose: Decide where to focus next month's marketing budget
Questions:
1. Which product category has the highest sales?
2. Which category has the highest/lowest month-over-month growth rate?
3. Are there any unusual patterns worth noting?
4. Which category should we focus on next month and why?
Output format:
- Answer each question by number
- Bold key figures
- End with 3-line action recommendations
Ready-to-Use Template
Analyze the following data.
[Data]
{{data or table}}
Analysis purpose: {{what decision will this analysis inform}}
What I want to know:
1. {{question 1}}
2. {{question 2}}
3. {{question 3}}
Output format:
- {{table / bullet points / narrative}}
- Highlight key figures
- Include {{action recommendations / important caveats}} at the end
12. Learning / Concept Understanding
Vague Request
What is RAG?
What level of explanation? Elementary student level? Developer level? Theory only or with practical examples? You end up with Wikipedia-level explanations.
Request with Criteria
Explain RAG (Retrieval-Augmented Generation).
My background: I know Python well, understand basic ML concepts, but
I'm new to NLP and LLM
What I want:
1. One-sentence definition
2. Why it's needed (why LLM alone isn't enough)
3. How it actually works (without code, using descriptions)
4. 2 real-world use cases
5. What I should learn first to understand this
Use easy analogies. When using difficult terms, add simple explanations in parentheses each time.
Ready-to-Use Template
Explain {{concept/technology/topic}}.
My background: {{current level and relevant knowledge}}
What I want:
- Explanation depth: {{beginner / intermediate / advanced}}
- Format: {{analogy-focused / example-focused / diagram (text) / code included}}
- Length: {{brief / medium / detailed}}
Include:
- {{e.g., real-world use cases / analogies / limitations / comparison with other concepts}}
Avoid:
- {{e.g., overly academic language / code examples / advanced concepts}}
13. Interview Preparation
Vague Request
Tell me interview questions.
What company? What position? What stage? Without this, you get a generic list of questions for any interview.
Request with Criteria
Help me prepare for an interview.
Company: Naver (200+ person IT company)
Position: Backend Developer (Python, FastAPI, 3 years experience)
Interview stage: 2nd technical interview (video, 1 hour)
My strengths:
- 2 FastAPI projects
- Redis caching optimization experience
- Experience in team with code review culture
Concerns:
- Weak at algorithm coding tests
- No large-scale system design experience
Request:
1. 10 questions likely to come up in this interview
2. Answer framework for each question (STAR method, etc.)
3. How to naturally highlight my strengths
4. How to honestly but positively answer about my concerns
14. Meeting Notes / Organization
Vague Request
Organize this transcript.
What format? Who will read it? What to focus on? AI might summarize narratively or cut too much, missing important decisions.
Request with Criteria
Organize the following meeting into meeting notes format.
[Transcript or notes]
{{content}}
Meeting type: Weekly sprint meeting
Attendees: 1 PM, 3 developers, 1 designer
Purpose: Share last week's progress, decide this week's priorities
Output format:
- Date/time/attendees header
- Main discussion items (by item, 2-3 lines each)
- Decisions made (specific)
- Action items (assignee / task / deadline in table format)
- Next meeting date
Exclude: Casual talk, repeated content, unresolved items
Language: Formal, concise
15. Presentation / Presentation Materials
Vague Request
Make presentation materials about AI.
How many slides? Who's the audience? What message? AI will give you content without story structure.
Request with Criteria
Create presentation structure and key content.
Topic: Proposal to introduce AI coding tools to our team
Audience: Non-developer executives (3 people)
Time: 15 minutes (5 minutes Q&A separate)
Goal: Get approval for adoption (including budget approval)
Key message: "AI tools can improve development speed by 40%"
Structure request:
- 10-12 slides
- Slide title + 2-3 key points each
- Where to place data/figures to persuade executives
- Anticipated objections and responses
What executives care about: Cost-benefit, security risk, learning burden
16. Proofreading / Feedback
Vague Request
Proofread this.
Just spelling? Change writing style? Check logic flow? Without direction, no clear standard for correction.
Request with Criteria
Proofread the following text.
[Original]
{{text}}
Proofreading perspectives (all that apply):
- Spelling / grammar errors
- Awkward sentences → natural
- Logic flow interruptions
Should maintain:
- My writing style and tone (don't change too much)
- Core argument and direction
Format:
- Full corrected text
- List corrected parts as [Corrected: original → corrected]
- Brief explanation for each correction
Common Standards for Good Prompts
A good prompt isn't about fancy sentences.
It's about clearly placing before AI what you want, why you need it, and what conditions must be met. The prompt changes when situations change. But there are recurring standards.
You provide context.
You state the purpose.
You define the format.
You set constraints.
You clarify what you don't want.
Asking AI well isn't about using more words.
It's about reshaping vague requests into forms that can be handled.