“I thought I asked a decent question, so why does ChatGPT keep giving me generic answers?”
Sound familiar? I’ve been there too. Then it clicked—the problem wasn’t ChatGPT. It was how I was asking.
In the AI world, this is called prompt engineering. Sounds fancy, but it’s really just “how to get AI to do what you want.” By the end of this post, you’ll be getting noticeably better results from ChatGPT.
1. Turn ChatGPT Into a 10-Year Veteran With One Line
Ask ChatGPT a plain question, and you’ll get a Wikipedia-style answer. But give it a role, and everything changes.
Here’s the difference:
Without a role:
“Tell me about performance marketing” → Generic definitions, textbook explanations
With a role:
“You are a performance marketing expert with 10 years of experience. Explain the core concepts of performance marketing to a new hire who knows nothing about marketing. Use real examples from actual ad campaigns.” → Practical insights, real-world examples, actionable advice
Copy and use this template:
You are an expert in [field] with [N] years of experience.
Explain [topic] to [target audience].
Focus on practical insights from real-world experience, not textbook theory.
Pro tip: Adding specific years of experience makes a difference. “Senior developer with 15 years of experience” gets you deeper answers than just “expert.”
2. ChatGPT Doesn’t Know Your Situation – Tell It
ChatGPT is smart, but it has no idea why you’re asking or what your context is. Give it background, and you’ll get tailored answers.
Without context:
“Help me write an email” → What kind of email? To whom? For what purpose?
With context:
“I’m a marketing manager at a tech startup. We’re launching a new product next week and I need to invite existing customers to the launch event. The email should highlight the 30% discount for attendees. Keep the tone friendly but professional.”
Context checklist (include these and watch the difference):
- Who you are (role, job)
- Why you need this (purpose)
- Who will see the result (audience)
- Any constraints (length, tone, format)
3. Vague Questions Get Vague Answers
This is the golden rule: specific input = specific output.
Here’s a quick comparison:
| Instead of this | Try this |
|---|---|
| “Keep it short” | “Around 500 words” |
| “Give me a few” | “Give me 5” |
| “For people” | “For mid-level engineers with 3-5 years of experience” |
| “Organize it” | “Format as a table” |
| “Make it good” | “Use a conversational but professional tone” |
Example prompt:
Write a blog post about 2025 digital marketing trends.
- Length: around 1,500 words
- Focus: SEO and social media marketing
- Audience: marketing professionals at mid-sized companies
- Tone: professional but easy to read
- Format: use H2 subheadings, 3-4 sections
The moment you add numbers and format specs, answer quality jumps significantly.
4. When Words Fail, Show an Example
Sometimes it’s hard to explain what you want. In that case, just show an example. This is called Few-shot Prompting.
Use it like this:
Write product descriptions following this example:
[Example]
Product: Wireless earbuds
Description: 🎧 Your daily commute companion. 8-hour battery life with IPX5 water resistance. Ergonomic design means all-day comfort without the ear fatigue.
[Now write]
Product: Portable humidifier
When you show an example, ChatGPT picks up on the tone, emoji usage, sentence length—everything. One good example beats a paragraph of explanation.
Tip: Save ad copy, emails, or reports you like and use them as reference examples later.
5. Get Copy-Paste Ready Output With Format Specs
Without format instructions, ChatGPT tends to ramble in paragraphs. Specify the format, and you get results you can use immediately.
Common format options:
- “As a table” → great for comparisons
- “As bullet points” → quick highlights
- “As a numbered list” → when order matters
- “In markdown” → ready for docs or blogs
- “As JSON” → for dev work
Example:
Compare these 5 programming languages in a table: Python, Java, JavaScript, Go, Rust
| Language | Primary Use | Learning Curve | 2025 Demand | Best For |
When you specify table headers upfront, you get exactly the columns you need.
6. For Complex Problems, Just Say “Step by Step”
For calculations or analysis, ChatGPT sometimes gets it wrong. Add “explain step by step” and accuracy improves dramatically.
This is called Chain of Thought (CoT) prompting. The idea is simple: instead of jumping to an answer and making mistakes, you force the model to think through each step.
Magic phrases:
- “Walk me through this step by step”
- “Show your reasoning”
- “Break this down one piece at a time”
- “Let’s think step by step” (this one’s particularly effective)
Example:
Help me allocate our marketing budget.
Total budget: $50,000
Channels: Google Ads, Instagram, YouTube, content marketing
Goal: Brand awareness for a skincare line targeting women 20-35
Analyze step by step and suggest budget allocation per channel.
Include reasoning and expected outcomes for each.
7. “Don’t Do This” – The Power of Constraints
Telling ChatGPT what not to do is just as important. Otherwise, answers can drift in unwanted directions.
Common constraints:
- “Avoid jargon”
- “Don’t mention specific brand names”
- “Skip the negative outlook”
- “No intro, get straight to the point”
- “No emojis”
Example:
Write an article about ethical AI use.
Constraints:
- Minimize technical jargon, keep it accessible
- Don't mention specific companies or products
- Balanced perspective, not doom-and-gloom
- Around 800 words
- Professional tone
8. You Don’t Need Perfect on the First Try – Iterate
Don’t expect a perfect answer right away. Refining through conversation actually produces better results.
ChatGPT remembers the conversation, so you can keep asking:
Refinement phrases:
- “Add more specific examples”
- “Make the tone more casual”
- “Expand the second section”
- “Summarize just the key points”
- “Rewrite this for a junior developer audience”
- “Make this part more engaging”
Pro tip: If the first answer is 70% there, that’s fine. Two or three rounds of refinement gets you where you need to be.
9. Tired of Repeating Yourself? Use Custom Instructions
“I’m a Python developer, please add comments to code, respond in a practical style…”
Typing this every time gets old. Set up Custom Instructions once and you’re done.
How to set it up:
- Click your profile icon (top right in ChatGPT)
- Select “Customize ChatGPT”
- Fill in both fields
First field – “What would you like ChatGPT to know about you?”:
- Backend developer, 5 years experience
- Primarily work with Python and PostgreSQL
- Building internal tools and APIs
- Prefer practical, no-fluff answers
Second field – “How would you like ChatGPT to respond?”:
- Include comments in code examples
- Skip unnecessary introductions, get to the point
- If unsure, say so honestly
- Use US English, include technical terms where appropriate
Set it once, applies to every conversation. Seriously convenient.
10. Ready-to-Use Prompt Templates
Enough theory. Here are copy-paste templates for common scenarios.
Blog Post Writing
[Role] You are a tech blogger with 10 years of experience.
[Topic] Write a blog post about {topic}.
[Audience] Tech-savvy readers who aren't necessarily experts
[Length] Around 2,000 words
[Format] Use H2 subheadings, 4-5 sections, include real examples
[Tone] Informative but conversational
[Note] Skip generic intros, get to the value quickly. No "In conclusion" section needed.
Email Writing
Write a business email for this situation:
Situation: {describe the situation}
From: {your role/title}
To: {recipient's role/relationship}
Purpose: {goal of the email}
Tone: {formal/casual/friendly professional}
Must include: {key points}
Provide 3 subject line options and the body text.
Reports and Proposals
[Background] {project context}
[Purpose] {what this document needs to achieve}
[Audience] {who will read this – executives/managers/peers}
Create a {document type} with these sections:
1. Current situation analysis
2. Problem areas / opportunities
3. Recommendations
4. Expected impact
5. Implementation plan
Length: {page count or word count}
Format: {tables, bullets, etc.}
Code Writing
[Language] {programming language}
[Goal] {what the code should do}
[Environment] {runtime, version, etc.}
[Constraints] {performance requirements, library restrictions}
Write the code with:
- Comments explaining key logic
- Error handling included
- Usage examples
Data Analysis
Analyze this data:
[Data]
{paste your data}
[Analysis requested]
1. Key trends or patterns
2. Outliers or notable points
3. Possible explanations
4. Recommended actions
Format results as tables and bullet points.
Interview / Presentation Prep
[Situation] {interview or presentation context}
[Question] "{expected question}"
[My background] {relevant experience}
Prepare an answer that:
- Uses STAR format (Situation-Task-Action-Result)
- Includes specific metrics or outcomes
- Fits in about 2 minutes
- Sounds natural when spoken
CO-STAR Framework (Universal Template)
For complex requests, this structure ensures you don’t miss anything.
[Context]
{background and situation}
[Objective]
{what you want to achieve}
[Style]
{writing style or format}
[Tone]
{voice and mood}
[Audience]
{who will see this}
[Response]
{desired output format and length}
Save these templates and pull them out when you need them. Start by filling in the blanks, and over time you’ll develop your own style.
TL;DR – Three Rules to Remember
Here’s the short version:
Three principles for better ChatGPT results:
- Be specific – include numbers, formats, and audience
- Give context – explain why you need it and who it’s for
- Iterate – don’t expect perfect on the first try
Prompting is a skill that improves with practice. Try these templates, tweak them, and build your own patterns.
Questions? Drop them in the comments.
Want to go deeper?