Not another “be specific” guide. These are the actual prompts and workflows I use every day — the ones that turned ChatGPT from a novelty into the most useful tool on my computer.
Why Most Prompting Advice Falls Flat
Every ChatGPT tips article tells you the same thing: “write clear prompts.” That’s like telling someone to “cook better food.” Technically true, completely useless.
The tricks that actually changed how I work weren’t abstract principles. They were specific prompts I could copy, paste, and immediately see the difference. So that’s what this post is — ten things I do on a regular basis, with the exact wording I type into the chat window.
Some of these are simple. A few took me months to figure out. All of them saved me real time once they became habits. I’m sharing the prompts verbatim so you can test them today, not next week after you’ve “optimized your workflow.”
Writing and Editing
1. The “write it wrong first” trick. Instead of asking ChatGPT to write something from scratch (where it defaults to generic corporate tone), I write a terrible first draft myself and paste it in. The prompt looks like this:
The output is dramatically better than asking “write a blog post about X.” The model is a much better editor than it is a writer from zero. Your messy draft gives it something real to work with instead of pulling from its default template. I’ve used this for everything from investor updates to newsletter intros, and the quality jump is noticeable every time.
2. The ban list. Large language models have a set of crutch phrases they reach for constantly. I keep a short ban list in my custom instructions that eliminates them permanently:
This single instruction changed the quality of every ChatGPT conversation I have. The writing comes out less robotic, less corporate, and more like something a human actually wrote. Add your own banned phrases as you notice them — everyone’s list grows over time.
3. Perspective-based editing. When I need to pressure-test a piece of writing — a pitch email, a proposal, a product description — I use this:
This is far more useful than “give me feedback.” By forcing the AI into a skeptical reader’s mindset, you get the kind of honest critique that’s actually hard to get from colleagues. I ran a cold sales email through this prompt last month and it flagged two assumptions I’d buried in the second paragraph that the prospect wouldn’t share. Rewrote those lines and got a reply the same day.
Data and Analysis
4. The spreadsheet whisperer. I paste messy data directly into ChatGPT and ask it to find what I’d miss staring at rows. Here’s a prompt I used recently with a client’s sales numbers:
The key phrase is “don’t just summarize.” Without it, you get a boring recap of what you already see. With it, you get analysis that actually surfaces insights. This works with any structured data — survey responses, website analytics, expense reports. Paste the raw numbers, tell it what you don’t want, and let it dig.
5. The meeting notes extractor. After any meeting, I paste the transcript (or my rough notes) and use this:
Category four is the one that saves me. The model is surprisingly good at catching the “we should probably do that” statements that aren’t real commitments but will cause problems if nobody follows up. I paste the output into our team channel within five minutes of every call ending.
Daily Productivity
6. The Monday morning planner. Every Monday I paste my calendar export and use this prompt. It takes 30 seconds and replaces 20 minutes of staring at Google Calendar.
The “context switching costs” part is what makes this useful. ChatGPT will notice when you have a creative meeting sandwiched between two analytical sessions and suggest rearranging the order. Small thing, but it compounds over weeks when you stop losing 15 minutes of mental warmup between mismatched tasks.
7. Scheduled tasks as a personal assistant. This is the ChatGPT feature most people still don’t know exists. You can schedule recurring prompts that run automatically using Scheduled Tasks (available on Plus and higher plans). Here’s what I have running:
- Every Monday 8 AM: “Search for the top 5 news stories in [my industry] from the past week. Give me a one-paragraph summary of each and rate how relevant it is to my work on a 1-5 scale.”
- Every Friday 4 PM: “Remind me to do my weekly review. Ask me these three questions: What went well? What didn’t? What’s the one thing I should prioritize next week?”
- 1st of each month: “It’s the first of the month. Remind me to review my subscription charges and check my budget tracker.”
This turns ChatGPT from something you remember to use into something that reaches out to you. The Monday briefing alone is worth the subscription cost — I stopped missing industry news that my competitors were already acting on.
Learning, Debugging, and the Self-Critique Loop
8. The “explain it back to me” trick. When I’m debugging or reviewing code I wrote weeks ago, I paste it in and ask:
This works for any technical material, not just code. Pasting in a legal contract with “explain each clause in plain language and flag anything unusual” has saved me from signing things I didn’t fully understand. The trick is asking it to flag the non-obvious parts — that’s where the real value hides, because those are exactly the things you’d skip over reading on your own.
9. The self-critique loop. When ChatGPT gives you a response and it feels “fine but not great,” don’t just rephrase your question. Instead, send this as a follow-up:
This works because it forces the model to evaluate its own output against specific criteria. The rewritten version is almost always significantly better — it fills gaps, removes hedge language, and adds concrete details that were missing the first time. I use this at least once per conversation now, usually after the first draft of anything important.
10. Build your custom instructions like a style guide. Every trick above works better when you set up Custom Instructions properly. Most people write something vague like “I’m a marketer, be helpful.” Here’s what mine actually looks like:
How to respond: Use short paragraphs. Lead with the answer, then explain. No corporate buzzwords — use the ban list. When I ask for help with writing, edit my draft rather than rewriting from scratch unless I say otherwise. For technical questions, include code examples. For strategy questions, give me a framework I can reuse, not a one-time answer.
▲ Key settings worth customizing in your instructions:
- Default output format — bullet points, short paragraphs, or structured sections
- Tone and vocabulary — the ban list plus preferred style
- Your expertise level — so the model doesn’t over-explain things you already know
- How to handle uncertainty — “say you don’t know rather than guessing”
Once these are set, every ChatGPT conversation starts from a better baseline. You stop repeating yourself. The tool already knows your preferences, your domain, and your communication style. It’s the difference between working with a new temp every day and working with an assistant who knows how you operate.
These ten ChatGPT tricks are not about prompting theory. They are specific workflows you can test today and measure the difference tomorrow. Start with whichever one matches your biggest daily friction point, run it for a week, and see what sticks.
Frequently Asked Questions
Nine out of ten work on the free tier. The exception is Scheduled Tasks (#7), which requires a Plus subscription or higher. Everything else — custom instructions, the ban list, the self-critique loop, data analysis prompts — works identically on the free plan. The free version uses GPT-4o mini, which handles structured prompts just as well for these use cases.
Two things make the biggest difference. First, set a ban list in your custom instructions to block overused phrases (“delve,” “streamline,” “it’s important to note,” etc.). Second, give it your own writing as a reference rather than asking it to write from scratch. When ChatGPT edits your draft instead of generating from zero, it naturally adapts to your voice instead of falling back on its default corporate tone.
Custom instructions (#10). It takes five minutes to set up and immediately improves every conversation you have going forward. Add your role, your preferred output style, and a short ban list. Everything else in this guide works better once that foundation is in place. After that, try the self-critique loop (#9) — it’s the fastest way to get noticeably better outputs without changing how you prompt.