I Thought ChatGPT Was Broken. Turns Out I Was Using It Wrong.
When I first started using ChatGPT to draft blog posts, I kept running into the same wall: the results felt off. Too generic. It seemed like the tool just didn't "get it."
Then, while setting up my first Project inside ChatGPT, I noticed something I'd ignored: a box labeled "Instructions."
That tiny detail led me to discover how prompts actually work — and a technique called in-context learning. Once I started giving ChatGPT a clear role, background, and specific direction, the quality of results jumped. I went from "this isn't working" to "this just saved me hours."
Here's how you can do the same.
Why Your Prompts Aren't Working
Let's get one thing clear: ChatGPT isn't actually "intelligent." It's more like a hyper-capable assistant who has read everything online but has no idea what's relevant to your situation unless you explain it.
The Golden Rule of Prompting: Specificity + Context = Quality Responses
Weak vs. Strong: A Real Example
Bad prompt: "Help me with marketing."
Better prompt: "Act as a marketing consultant for small businesses. I run a local plumbing company and want to create a 30-day social media content calendar focused on home maintenance tips that will build trust with homeowners. Give me 10 post ideas with brief descriptions."
The second prompt gives ChatGPT the role, audience, goal, and format — everything it needs to do great work.
The 9 Essential Prompting Techniques
1. Be Ruthlessly Specific
Don't just say "write an email." Say what kind, for what purpose, and to whom.
Try: "Write a follow-up email to a potential client who viewed our proposal but hasn't responded in two weeks."
2. Provide Rich Context
Give ChatGPT the why behind your request: your audience, your goals, and any constraints.
3. Assign a Role
Tell ChatGPT who it is in the conversation. It will adjust its tone and expertise accordingly.
Try: "Act as an experienced business coach who works with solopreneurs."
4. Control the Length
Be clear about what you want. "Summarize this in three bullet points." Or "Create a 500-word blog post."
5. Break Down Complex Tasks (Prompt Chaining)
Instead of one giant request, split big tasks into step-by-step prompts.
- Step 1: "Outline my client onboarding process."
- Step 2: "Now draft the welcome email for step 1."
- Step 3: "Create a checklist based on the outline."
6. Use Examples (Few-Shot Prompting)
Want it to sound like you? Give it examples. "Write posts like this: [example]. Now give me three more."
7. Ask for Step-by-Step Thinking
When you need logic or strategy, tell it to think out loud.
Try: "Walk me through whether I should hire a full-time assistant or keep outsourcing. Consider costs, benefits, and long-term ROI."
8. Refine and Iterate
Never settle for the first draft. Ask it to adjust tone, add examples, or rewrite for clarity.
"Make this more conversational." / "Add specific examples." / "Rewrite this for someone new to the industry."
9. Always Verify Important Info
ChatGPT isn't a fact-checker. Double-check dates and statistics, legal or financial information, and technical specifications.
Choosing the Right ChatGPT Model
| Model | Best For |
|---|---|
| GPT-5.5 | The default for most tasks — writing, customer support, live chat, images. Fast and capable across the board. |
| GPT-5.4 Thinking | Complex writing, nuanced analysis, and emotional tone. Use it when depth and precision really matter. |
| Mini model | Fast and lightweight. Great for quick drafts, social captions, and simple questions. |
| o3 | OpenAI's top reasoning model. Ideal for strategic planning, logic-heavy tasks, and complex problem solving. |
FIVE75 tip: Start with GPT-5.5 for most daily work. Switch to GPT-5.4 Thinking when the task requires deeper analysis or more precise output.
Your Action Plan for This Week
- Rewrite one prompt — take a task you struggled with and rewrite it using Specificity + Context.
- Experiment with roles — try giving ChatGPT two or three different roles for the same task.
- Break a big task into steps — use prompt chaining to turn a complex task into a simple sequence.
- Compare models — run the same prompt through GPT-5.5 and GPT-5.4 Thinking. Notice the difference in depth and nuance.