How I Actually Use ChatGPT: My Daily Workflow

I've been using ChatGPT almost every day for 18 months. Not dabbling, but actually integrating it into real work. Writing, coding, research, thinking through problems.

Most "how to use ChatGPT" articles give you generic tips. "Be specific!" "Give context!" Sure. But what does that actually look like in practice?

This is my real workflow. The actual prompts I use, the patterns I've developed, and the mistakes I've learned to avoid. If you're already past the basics and want to see how someone uses ChatGPT professionally, this is for you.

My Setup: ChatGPT Plus + Custom Instructions

First, the basics. I pay for ChatGPT Plus ($20/month). The free tier is too limited for serious work: slower responses, usage caps, no GPT-4. If you're using ChatGPT for actual productivity, the subscription pays for itself in the first week.

More importantly: I've configured Custom Instructions. This is the most underrated feature. Here's roughly what mine say:

About me:

I'm a developer and writer. I work on web applications (React, TypeScript, Node.js) and write articles about productivity and AI. I prefer concise, direct communication. I value accuracy over speed. I'd rather you say "I'm not sure" than guess.

How I want responses:

Be direct and concise. Skip the preamble ("Great question!"). When I ask for code, give me production-quality code with proper error handling. For writing help, match my voice: conversational but substantive. If my request is ambiguous, ask one clarifying question before proceeding.

These instructions shape every conversation. ChatGPT stops adding fluff, gives me better code by default, and asks for clarification instead of guessing. It took me months to realize how much Custom Instructions improve the experience.

Writing Workflow: Not What You'd Expect

I don't use ChatGPT to write for me. I use it to think with me.

When I'm planning a new piece, I don't say "write an article about X." That produces generic AI slop. Instead:

I'm writing an article about [topic]. My main argument is [thesis].

Before I start writing, help me stress-test this idea:
1. What's the strongest counterargument?
2. What am I probably overlooking?
3. What evidence would make this more convincing?

Don't write the article. Just help me think through it.

This is the key insight: ChatGPT is a thinking partner, not a writing machine. The output I want isn't text. It's better thinking.

After I write a draft (myself, not AI), I use ChatGPT to edit:

Here's a paragraph from my draft. It feels flabby. Identify:
1. Words I can cut without losing meaning
2. Sentences that repeat the same idea
3. Where I'm telling instead of showing

Don't rewrite it. Just point out the problems so I can fix them.

[paste paragraph]

Notice the pattern: I ask ChatGPT to identify problems, not fix them. This keeps my voice intact while benefiting from a second perspective.

One area where I do use ChatGPT for generation: headlines. I'm too close to my own work to see it fresh.

Here's my article about [topic]. The main takeaway is [key point].

Give me 10 headline options. I want:
- Specific, not generic ("7 Tools" not "Some Tools")
- Benefit-focused when possible
- No clickbait or exaggeration

[paste first 500 words]

I never use the headlines verbatim, but they spark ideas I wouldn't have found alone.

Coding Workflow: Where ChatGPT Shines

Coding is where ChatGPT has genuinely transformed my work. Not for writing entire applications, but for specific, targeted tasks.

Debugging is my most common use case. When I hit an error I don't immediately understand:

I'm getting this error in my React app:

[paste error message]

Here's the relevant code:

[paste code]

What's causing this, and what's the fix? Explain why the error happens so I understand it, not just how to make it go away.

The "explain why" part is crucial. I want to learn, not just copy-paste a fix. This approach has taught me more about React edge cases than any tutorial.

For well-defined, isolated functions, ChatGPT is excellent:

Write a TypeScript function that:
- Takes an array of objects with 'date' and 'value' properties
- Groups them by week
- Returns the sum of values for each week

Include:
- Proper TypeScript types
- Handle empty arrays
- Handle invalid dates gracefully
- JSDoc comments

I'm specific about requirements upfront because vague prompts produce vague code. The more constraints I give, the better the output.

Before committing complex code, I ask for a review:

Review this function for:
1. Potential bugs or edge cases I'm missing
2. Performance issues
3. Readability improvements

Be critical. I want to catch problems before they ship.

[paste code]

ChatGPT won't catch everything a human reviewer would, but it catches enough to be valuable. It's especially good at spotting off-by-one errors and null pointer issues.

When I'm picking up a new library or framework:

I'm learning [technology]. I already know [relevant background].

Explain the core mental model I need. Not a tutorial, just the key concepts that will make everything else click.

Then give me one small, practical example that demonstrates those concepts.

This approach is faster than reading documentation because it's targeted to what I already know. ChatGPT bridges the gap between my current knowledge and the new thing.

Thinking Workflow: The Underrated Use Case

This is where most people underuse ChatGPT. It's a surprisingly good thinking partner.

For significant decisions:

I'm deciding whether to [decision]. Here's the context:

[brief background]

Help me think through this:
1. What are the second-order consequences I might be missing?
2. What would I need to believe for each option to be correct?
3. What's the reversibility of each choice?

Don't tell me what to do. Help me see the full picture.

This pairs well with keeping a decision journal. ChatGPT helps me think more systematically before I commit to a choice.

Often the real problem isn't the one I think I'm solving:

I'm trying to solve [problem as I currently see it].

Before I jump to solutions, help me question the framing:
1. Am I solving the right problem?
2. What assumptions am I making?
3. Is there a simpler version of this problem I should solve first?

This has saved me weeks of work on multiple occasions. Problem framing is an underrated skill, and ChatGPT is good at pressure-testing my framing.

Sometimes I use ChatGPT as a rubber duck, but a smart one:

I'm going to explain my reasoning for [decision/approach]. 

Listen, then poke holes in my logic. Where am I being inconsistent? Where are the weak points?

[explain reasoning]

Articulating my thinking to an AI often reveals gaps I couldn't see internally. The act of explaining is itself valuable.

What I Don't Use ChatGPT For

Just as important is knowing where ChatGPT isn't helpful:

• First drafts of important writing. AI-written text lacks voice. I write first, then use AI to edit.
• Anything requiring current information. ChatGPT's knowledge cutoff matters. For recent events, libraries, or news, I use other sources.
• Final decisions. ChatGPT helps me think, but I make the call. It's an advisor, not a decider.
• Tasks I should learn myself. If I'll do something repeatedly, I learn it properly rather than outsourcing to AI.

Patterns That Took Me Months to Learn

Ask for critique, not creation. "What's wrong with this?" produces better output than "Write this for me." ChatGPT is better at improving than creating from scratch.

Request explanations, not just answers. Adding "explain why" or "help me understand" produces responses that actually build my knowledge instead of creating dependency.

Give examples of what you want. "Write it like this example: [paste example]" is more effective than describing the style you want. Show, don't tell.

Iterate in the same conversation. Follow-up prompts in the same chat are better than starting fresh. ChatGPT has context about what you're trying to do.

Save prompts that work. When I find a prompt structure that produces good results, I save it. My notes app has a "ChatGPT prompts" folder with templates for common tasks.

The mindset underneath the workflow

The workflow above only works because of how I frame the tool itself. This section was originally a separate article; it belongs here, next to the prompts it explains.

The Answer Machine Trap

Here's how most people use ChatGPT:

• "Write me an email to my boss about the project delay."
• "Give me 10 blog post ideas about productivity."
• "What's the best way to learn Python?"
• "Summarize this article for me."

These prompts treat AI as a vending machine. Insert question, receive answer, done. It feels efficient. But there's a hidden cost: you're skipping the thinking that makes you smarter.

When you ask ChatGPT to write your email, you don't wrestle with how to frame the message. When you ask for blog ideas, you don't exercise your creative muscles. When you ask for the "best" way to learn something, you outsource judgment to a model that doesn't know your context.

The result? You get outputs without developing the mental models that create real expertise. You become dependent on the tool instead of enhanced by it.

Five Ways to Use ChatGPT as a Thinking Partner

First: challenge your assumptions. Before making a decision, ask ChatGPT to argue against your position. "I'm planning to [decision]. Play devil's advocate. What could go wrong? What am I not seeing?"

This isn't about getting the "right" answer. It's about surfacing blind spots you might have missed. I do this before major decisions, and it's saved me from several mistakes. For more on structured decision-making, check out The Decision Journal.

Second: explore multiple perspectives. When you're stuck, ask for viewpoints you haven't considered. "I'm trying to solve [problem]. How would a designer approach this? A behavioral economist? A minimalist?"

The goal isn't to blindly follow any perspective. It's to expand your solution space. Often the best answer comes from synthesizing multiple viewpoints into something new.

Third: rubber duck your ideas. Programmers use a technique called "rubber duck debugging": explaining code to a rubber duck to find bugs. ChatGPT is an excellent rubber duck.

"Here's my plan for [project]. I'll walk you through it step by step. After each step, ask me one clarifying question."

The act of explaining forces clarity. The questions reveal gaps. You end up with a better plan because you did the thinking, not because AI did it for you.

Fourth: build on your drafts. Instead of asking ChatGPT to write from scratch, write a rough draft yourself first. Then use AI to improve it.

"Here's my draft of [content]. What's unclear? What could be stronger? Give me specific suggestions, but don't rewrite it. I want to understand why."

This approach keeps you in the driver's seat. You learn from the feedback instead of just consuming the output. Your writing actually improves over time.

Fifth: learn the "why" behind answers. When ChatGPT gives you information, don't stop there. Dig into the reasoning.

"You recommended [approach]. Walk me through the logic. What principles is this based on? When would this advice be wrong?"

Understanding the "why" builds transferable knowledge. Next time you face a similar situation, you'll have a mental model, not just a memorized answer.

A Simple Framework to Start

Next time you open ChatGPT, pause before typing. Ask yourself:

• Am I trying to skip thinking, or enhance it?
• What do I already know or believe about this?
• What specific help would make my thinking better?

Then craft a prompt that treats AI as a collaborator, not a replacement. Share your context, your current thinking, your constraints. Ask for perspectives, trade-offs, challenges. Not just answers.

It takes a bit more effort. But you'll get better outputs, develop stronger judgment, and actually learn something in the process.

That's the difference between using AI as a crutch and using it as a lever. One makes you weaker. The other makes you stronger.

Want to sharpen your thinking in other areas? Check out Learning How to Learn for science-backed techniques that compound over time, or explore 5 AI Tools Every Entrepreneur Should Use for practical applications beyond ChatGPT.

The Meta-Lesson

After 18 months, here's what I've learned: ChatGPT is not a shortcut. It's a tool that amplifies your existing skills.

If you're a mediocre thinker, ChatGPT won't make you a good one. But if you already think carefully, it helps you think faster and more systematically.

If you write poorly, ChatGPT will produce poor writing for you. But if you write well, it helps you refine your work more efficiently.

The people getting the most value from AI aren't using it as a replacement for skill, they're using it to build on skills they've already developed.

Don't ask "how do I get ChatGPT to do my work?" Ask "how do I use ChatGPT to do my work better?"

That shift in framing changes everything about how you approach the tool. And it's the difference between getting generic AI output and getting genuinely useful assistance.

Start with one workflow from this article. Try it for a week. Adapt it to your needs. That's how you build a real AI workflow: through actual use, not theory.