The first time I used ChatGPT, I asked it: “Write me a blog post about productivity.” It gave me 800 generic words that read like it was written by a robot (which, technically, it was). I was underwhelmed. Tried a few more times with similar generic prompts and got similar mediocre results.
Then something clicked. I stopped asking ChatGPT to “write” things and started using it like a thinking partner. I gave it context, constraints, specific formats, and actual examples of what I wanted. The results went from “okay to publish after heavy editing” to “I can use this with minimal changes.”
That’s when I realized: most people are using ChatGPT wrong. Not wrong exactly, but inefficiently. They’re getting 40% of what it could do because they’re not prompting it properly.
Over the past year, I’ve figured out what actually works. I use ChatGPT daily now—for content research, outlines, editing, brainstorming, even coding help. And I’ve learned that how you ask the question matters more than what tool you’re using.
Let me show you the tricks that actually changed my productivity.
The Problem: Why Generic Prompts Fail
Before I show you what works, let me explain why most ChatGPT prompts produce mediocre results.
ChatGPT is like a really smart assistant who’s never worked with you before. If you give them vague instructions, they’ll do something reasonable based on their general knowledge. But they don’t know your style, your audience, your constraints, or what “good” looks like for your specific situation.
“Write a blog post” means something different to me than it does to a financial advisor or a recipe blogger. But ChatGPT doesn’t know that. So it writes something generic that could apply to anyone.
The people getting incredible results from ChatGPT aren’t smarter. They’re just better at giving context.
Trick 1: Give It a Role (And Make It Specific)
Most people start with “ChatGPT, write…” Instead, start with “You are…”
Bad prompt: “Write an email to my boss asking for a raise.”
Better prompt: “You are an employee with 5 years at a tech company who has consistently exceeded targets and taken on leadership responsibilities. Write an email to my engineering manager (who values data and straightforwardness) asking for a 15% raise. The tone should be professional but warm, and the email should be 3 paragraphs max.”
The difference is specific role definition. I’m not just asking ChatGPT to write an email. I’m telling it exactly who’s writing, what context matters, and what constraints exist.
I use this constantly. “You are a conversion copywriter who specializes in SaaS products…” or “You are a technical writer who makes complex concepts simple for beginners…”
Real example from my work: I needed LinkedIn captions. Instead of “Write a LinkedIn caption about productivity,” I said: “You are a productivity consultant with a cynical sense of humor who works with solopreneurs. Write a LinkedIn caption (2-3 sentences max) that makes fun of productivity culture while subtly promoting better systems. Make it shareable but not preachy.”
The results went from generic motivational quotes to actually funny, original stuff that got engagement.
Trick 2: Use Examples, Not Descriptions
If you want specific output, show ChatGPT examples rather than describing what you want.
Bad approach: “Write bullet points about a productivity app in an informal, witty tone.”
Better approach: “Here are examples of the tone I want:
- This app doesn’t make you productive. It just makes procrastination feel more intentional.
- Use it to track tasks. Ignore the motivational quotes—they’re there, but no one asked for them.
- The real feature is calendar integration. The rest is just pretty.
Now write 5 similar bullet points about Notion.”
Showing examples is incredibly powerful because ChatGPT understands patterns. When you describe “informal and witty,” everyone has a different mental image. When you show examples, there’s no ambiguity.
I’ve started keeping a “tone examples” document. Whenever I write something I really like, I save it. Then when I need ChatGPT to match that tone, I just paste an example.
Real example: I wanted product descriptions in a specific style—funny but informative, not trying too hard. I pasted three examples of product descriptions I loved, and ChatGPT matched that style almost perfectly. Without examples, it defaulted to generic marketing language.
Trick 3: Define Your Constraints Upfront
ChatGPT is constraint-focused. Give it good ones and it performs better.
Instead of just asking for content, tell it:
- Length (300-400 words, not longer)
- Format (3 sections with subheadings, not a wall of text)
- Tone (professional but not stuffy)
- What to avoid (no clichés, no jargon, no “revolutionary”)
- Who it’s for (beginners, not experts)
“Write a guide to email marketing for beginners (400-500 words, avoid industry jargon, 3-4 practical tips only, conversational tone, no hype language like ‘game-changer’).”
Notice how specific that is? ChatGPT now has guardrails. It knows what you don’t want, not just what you do.
I’ve noticed that content with explicit constraints is almost always better than content with vague descriptions. It’s like the constraints focus ChatGPT’s attention.
Trick 4: Ask for Iteration, Not Perfection
Most people ask ChatGPT for something once and take the result. That’s leaving quality on the table.
Instead, treat it like collaborating with a human. Get the first draft, then iterate.
“Now make it more sarcastic.” “Remove the paragraph about X, it’s not relevant.” “Make the first paragraph punchier—I want people to actually read this.” “Add a sentence about Y because my readers care about that.”
I usually go through 2-3 iterations before I’m happy. Each time I’m giving ChatGPT more information about what I want.
Real workflow: I asked for a sales email. First version was okay but formal. Second: “Make it more like we’re friends talking, not a business email.” Better. Third: “Add a specific example of the problem we solve—people need to see they’ve experienced this exact thing.” That version actually converted.
The magic isn’t in the first response. It’s in the conversation.
Trick 5: Use “Think Step-by-Step” For Complex Tasks
When you want ChatGPT to actually reason through something, explicitly ask it to think out loud.
Bad: “Should I start a new blog or focus on my existing one?”
Better: “I have two options: start a new blog about X or grow my existing blog about Y. Walk me through the pros and cons of each. Think step-by-step about traffic, monetization timeline, audience overlap, and effort required. Then give me a recommendation.”
By asking it to think step-by-step, ChatGPT actually breaks down the reasoning rather than just giving you a quick answer. You get to see its logic, which helps you understand the recommendation better (and spot if it’s missing something).
I use this for anything strategic or complex. Business decisions, content strategy, technical problems. “Think step-by-step” forces actual reasoning instead of pattern matching.
Trick 6: Ask for Multiple Options, Then Pick
Instead of asking ChatGPT for one answer, ask for 3-5 options and pick the best.
“Give me 5 different angles for a blog post about AI tools for content creators. List them as:
- [Angle]
- [Angle] etc.
Then tell me which one you think would drive the most traffic, and why.”
I always find that option 2 or 3 is better than option 1, but I wouldn’t have thought of it without seeing alternatives. Getting multiple options is like brainstorming with someone smart.
Real example: I asked for 5 different subject lines for an email. The one ChatGPT suggested as best wasn’t my favorite, but option 4 sparked an idea for a better one. I wouldn’t have gotten there with just one suggestion.
Trick 7: Reference Your Own Content
ChatGPT doesn’t know what you’ve written. So if you want it to match your style or avoid repeating yourself, you have to show it.
“Here’s an article I published about X. Now write a follow-up article about Y that builds on this but doesn’t repeat the same points. Keep the same conversational tone.”
Paste your previous article and ChatGPT will reference it, avoid repetition, and maintain your voice better.
This has saved me hours of editing. Instead of ChatGPT writing something that’s 60% repetitive of my previous work, it deliberately builds on it and stays fresh.
I do this especially for series content. “Here’s article 1, 2, and 3 in this series. Now write article 4 that assumes readers know this background but doesn’t repeat the basics.”
Trick 8: Give It Information It Doesn’t Have (Your Data)
ChatGPT’s training data has a cutoff. So for anything recent, you need to feed it the information.
“It’s June 2026. Here’s information about the latest Google search changes: [paste relevant info]. Now write about how this affects small blogs. Use the specific information I provided, not general knowledge.”
I do this with:
- Recent industry changes
- Personal data about my business
- Specific research or studies
- Current pricing or features
- Niche information it wouldn’t know
By feeding ChatGPT the exact information you want it to work with, you get more accurate, more relevant results.
Real example: I wanted ChatGPT to comment on recent ad network changes. I pasted three recent articles about it and asked ChatGPT to synthesize them into a blog post. The result was accurate and current instead of generic.
Trick 9: Ask It to Point Out What’s Missing
ChatGPT is great at generating content, but sometimes it misses what would actually be useful.
After getting a first draft, ask: “What would make this better? What important point am I missing? What would actually help someone reading this?”
ChatGPT will often say something like “You should probably include a section about X because readers care about that” or “This would be stronger with specific examples.”
Then you ask it to add those things.
I’ve found that ChatGPT is better at critiquing its own work when you explicitly ask it to. It’s like it has ideas but doesn’t always include them in the first pass unless prompted.
Trick 10: Save Your Best Prompts
This is the simplest trick but it compounds over time.
Whenever you write a prompt that produces really good results, save it. Create a document or a folder of prompts that work for your specific use cases.
Over time, you build a personalized prompt library. Need an email? Use your email template prompt. Need product copy? Use your proven copy prompt. Need research? Use your research prompt.
I have about 15 prompts I use constantly. Each one is refined from repeated use. They’re more powerful than generic prompts because they’re tailored to exactly what I do.
My most-used prompts:
- Content outline template (for any article)
- Email template (for promotional, informational, etc.)
- Research summarizer (for analyzing competitor sites)
- Social media caption framework
- Product description template
- Technical explanation simplifier
Having these ready saves me probably 3-4 hours per week.
The Real Skill: Context, Not Prompts
Here’s what I’ve realized: the best prompt isn’t some magic formula. It’s providing enough context that ChatGPT understands what you actually need.
Most people treat ChatGPT like a search engine—quick question, quick answer. But it works better as a thinking partner. Give it background, constraints, examples, and your actual goals. Have a conversation. Iterate.
The people getting amazing results aren’t smarter than everyone else. They’re just being more deliberate about what they ask and how they ask it.
How to Get Started
Pick one prompt you use frequently (email, content outline, whatever) and rebuild it using these tricks:
- Start with a role: “You are…”
- Add constraints: “Length: X words, tone: Y, format: Z”
- Give examples: “Here’s what I mean by…”
- Add context: “This is for [audience] who [situation]”
- Iterate: “Now make it…”
Then save that prompt. Refine it over time. Use it repeatedly.
Don’t try to use all 10 tricks on every prompt—that’s overkill. But use 3-4 of them deliberately and you’ll get dramatically better results.
ChatGPT isn’t magic. You just have to talk to it properly.