There's a pattern I see constantly in how people talk about ChatGPT. It goes like this:
"I tried ChatGPT for [thing] and the output was mediocre. Feels overhyped."
When I ask what they typed, it's always a version of the same thing: a short, context-free sentence that vaguely describes what they want. Something like: "Write me a professional email to my client about the delay." Or "Give me some ideas for a YouTube channel." Or "Help me with marketing."
These prompts aren't prompts. They're wishes.
And ChatGPT responds to wishes with generic outputs, because that's all it has to work with.
The Fundamental Misunderstanding
Most people treat ChatGPT like a search engine — type a short query, get a usable result.
But ChatGPT isn't a search engine. It's a collaborative text interface. It responds to what you give it. If you give it nothing — no context, no audience, no constraints, no goal — it generates content that could apply to anyone and therefore applies to no one in particular.
This is the entire reason most people have mediocre experiences with it. Not the model's limitations. Their inputs.
Here's a direct comparison to show what I mean.
Bad prompt: "Write a LinkedIn post about productivity."
Output: Generic tips nobody needed.
Better prompt: "I run a solo consulting business and I'm writing a LinkedIn post for other independent consultants. I want to share a specific insight I had this week: I used to batch all my client calls on Tuesdays, and it always left me depleted for the rest of the day. I switched to distributing them across the week with protected deep-work mornings, and it changed how I feel by Friday completely. Write this as a first-person LinkedIn post in a conversational, direct tone. No fluff. Open with the problem, not with 'here's what I learned.' Under 200 words."
Output: A specific, personal, well-structured post that sounds like a real person — because it has real context to work with.
Same tool. Completely different output. The difference is the quality of the input.
What a Good Prompt Actually Contains
If you want usable output, your prompt needs four things:
1. Context — Who are you? What's the situation? What platform or format is this for?
2. Audience — Who is the output for? Not "general" — a specific type of person.
3. Constraints — What boundaries define a good output? Length, tone, format, things to avoid.
4. Examples or analogies — What does success look like? Paste an example you like, describe the style, name a reference.
None of these requirements are complex. They're just the information a real human collaborator would need before doing the work.
Three More Mistakes (With Fixes)
Mistake 1: Accepting the first output.
The correct response to an output that's not quite right is to tell ChatGPT what's not quite right. "This is too formal. Rewrite it more conversationally." "The opening is weak. Give me 5 alternative opening lines." ChatGPT holds the context of your conversation — you're directing a collaboration, not starting over.
Mistake 2: Using it to generate ideas from nothing.
"Give me business ideas" produces generic ideas with no relationship to your skills, situation, or market. Fix: "I have 10 hours per week, $500 to invest, skills in [X], and I want to generate $1,000/month within 6 months. I'm not interested in dropshipping or MLM. What are some realistic options?" Now it has something to work with.
Mistake 3: Not giving it your actual writing to learn from.
If you want ChatGPT to write in your voice, give it examples of your voice. "Here are three paragraphs I wrote: [paste]. Now write a new paragraph on [topic] that matches this style, tone, and voice exactly." This single technique changes output quality dramatically.
What ChatGPT Is Actually Good At
Thinking tools: Outlining, structuring arguments, steelmanning positions I disagree with, finding the holes in a plan.
Research acceleration: Summarizing complex topics, explaining technical concepts in plain language.
Writing infrastructure: First drafts, structural frameworks, headline options, subject line testing, reformatting content for different platforms.
Not good at: Original research (hallucination risk), real-time information, personal decisions with high stakes, and anything where your lived experience is the actual value.
The Reframe That Changes Everything
Stop thinking of ChatGPT as a tool that produces outputs. Start thinking of it as a thinking partner that requires information to think with.
Every time you're underwhelmed by ChatGPT, ask the question: what didn't I tell it? The answer is almost always the path to a better prompt and a better output.
The gap between "AI is overhyped" and "AI transformed how I work" usually isn't the model. It's the quality of the conversation you're willing to have with it.
I write about practical AI use, solo business building, and the tools I actually use week to week. Hit follow if that's your thing.
Top comments (0)