I Asked the Same Thing Two Ways — One Answer Was Useless, the Other Blew My Mind
See how rephrasing the same question to ChatGPT produces wildly different results. Real examples and the simple fix anyone can use.
Last week I wanted help writing an email to reschedule a meeting. I typed a quick question into ChatGPT, got a generic response, and almost gave up. Then I rephrased it — same request, different words — and the result was so good I used it word-for-word.
This happens more often than you'd think. The way you ask determines what you get back.
The First Attempt: Vague and Forgettable
Here's what I typed first:
prompt Help me write an email to reschedule a meeting.
ChatGPT gave me something like this:
"Dear [Name], I hope this email finds you well. I am writing to request a change to our scheduled meeting..."
Technically correct. Also boring, robotic, and something I'd need to completely rewrite anyway. It didn't know who I was emailing, why I was rescheduling, or what tone I needed.
The Second Attempt: Specific and Useful
I tried again with more context:
prompt I need to reschedule a meeting with my manager about my quarterly review. The original time was Thursday at 2pm but I have a dentist appointment I forgot about. I want to sound professional but not overly formal since we get along well. Suggest moving it to Friday morning instead. Keep it short — 3-4 sentences max.
The response was completely different. It mentioned the dentist appointment naturally, suggested Friday morning, kept the friendly-but-professional tone, and was exactly four sentences. I copied it, changed one word, and hit send.
Why the Difference Matters
AI tools like ChatGPT don't read your mind. They work with what you give them. The first prompt gave almost nothing — no context, no constraints, no personality. So the AI filled in the blanks with the most generic option possible.
The second prompt told the AI:
- Who I'm emailing (my manager)
- What it's about (quarterly review)
- Why I'm rescheduling (dentist appointment)
- The tone I want (professional but friendly)
- A specific alternative (Friday morning)
- A length limit (3-4 sentences)
More information in means better output out. It's that simple.
The Pattern I Now Use Every Time
After testing this with dozens of prompts, I found a formula that works:
Context + Task + Constraints = Good Results
- Context: Background information (who, what, why)
- Task: What you actually want (write, summarize, explain, compare)
- Constraints: Limits and preferences (length, tone, format, audience)
Here's another example. Bad prompt:
prompt Explain machine learning.
You'll get a textbook definition that might confuse more than help.
Better prompt:
prompt Explain machine learning to someone who has never touched code. Use an everyday example like Netflix recommendations. Keep it under 100 words and avoid technical jargon.
Now you'll get something you can actually use — maybe to explain the concept to a colleague or include in a presentation.
Three Quick Fixes for Any Prompt
If your AI responses feel generic, try adding one of these:
1. Add "for" + your situation
Instead of "Write a LinkedIn post," try "Write a LinkedIn post for a freelance graphic designer looking for clients in the food industry."
2. Add a format constraint
Instead of "Give me ideas," try "Give me 5 ideas as bullet points, each one sentence max."
3. Add a tone reference
Instead of "Write a bio," try "Write a bio that sounds confident but not arrogant — like someone who's good at their job but doesn't take themselves too seriously."
The Real Lesson Here
The AI didn't get smarter between my two attempts. I got better at asking.
This isn't about memorizing magic phrases or special tricks. It's about realizing that a prompt is a conversation starter. The more you give, the more relevant the response.
Next time you get a useless AI response, don't blame the tool. Look at your prompt first. Add one more sentence of context. Specify what you actually want. Set a constraint or two.
The difference between "meh" and "wow" is often just 15 extra words.
Did you like this?
Every Monday I send one email with a prompt or a tool I tried. No spam, unsubscribe with one click.
Subscribe to the newsletter