How to Actually Use ChatGPT (Not the Generic Tips Everyone Repeats)

TL;DR

ChatGPT works best when you stop treating it like Google and start treating it like a smart intern — give it enough context, check its work, and don't be afraid to push back. The free tier is plenty for most people. Custom Instructions and memory are the two features that make the biggest difference, and almost nobody sets them up.

I signed up for ChatGPT the week it launched in November 2022, mostly out of curiosity. Two hours later I realized I'd accidentally spent my whole evening asking it to explain Rust's borrow checker, draft emails I'd been putting off, and generate terrible poetry about my cat.

Since then I've used it nearly every day — for coding, writing, research, and occasionally settling arguments with friends. Here's what I've learned about actually getting value from it, beyond the "write better prompts" advice everyone gives.

What ChatGPT Actually Is (and What It Isn't)

ChatGPT is a language model built by OpenAI. It predicts what text should come next based on patterns it learned from a huge chunk of the internet. That's it. It doesn't think, doesn't understand, doesn't have opinions. It's pattern matching at a scale that feels like intelligence.

This matters because it explains why ChatGPT is simultaneously brilliant and dumb. It can write a working Python script in seconds but confidently tell you that 2 + 2 = 5 if the surrounding context makes that look like the most probable answer.

Treat it like a very eager, very knowledgeable intern who will never admit when they don't know something. You need to give clear instructions, provide context, and verify anything important.

Setting Up (5 Minutes, Tops)

Go to chat.openai.com, sign up with an email or Google account, and verify your email. That's it.

Free vs Plus ($20/month):

| Tier | What you get | Who it's for | |------|-------------|-------------| | Free | GPT-4o mini, limited messages, no file uploads | Casual use, learning, occasional tasks | | Plus | GPT-4o, higher message limits, file uploads, DALL-E, Custom GPTs | Daily users, professionals, anyone doing real work |

Start with the free tier. If you find yourself hitting the rate limit regularly or needing file uploads, upgrade. I used the free tier for six months before switching to Plus, and honestly — for most things, the free tier was fine.

Two settings worth changing right now:

Custom Instructions (Settings → Personalization → Custom Instructions): Tell ChatGPT about yourself — what you do, how you want it to respond, what tone you prefer. This single setting transforms the quality of responses more than any "prompt hack" I've seen.
Data controls (Settings → Data Controls → "Improve the model for everyone"): Turn this off if you're pasting sensitive code, client documents, or anything private. Takes five seconds.

Here's what my Custom Instructions look like:

I'm a software developer who writes technical blog posts and builds web apps. I prefer direct, concrete answers without fluff. Use code examples when relevant. Don't explain basic programming concepts unless I ask.

Adjust to fit what you do. The difference is night and day.

Prompts: The Thing Nobody Actually Explains

Most "prompt engineering" guides give you templates and formulas. I've never used a template in my life. What actually matters is understanding three things:

1. Context is everything

Bad prompt:

Write me a blog post about React hooks.

Good prompt:

I'm writing a tutorial for junior developers who've been using class components for a year. They're switching to hooks and are skeptical about the benefits. Explain useState and useEffect with before/after examples. Keep it practical, skip the theory. Target ~300 words.

The second prompt works because ChatGPT knows who it's writing for, why they care, what format you want, and what to avoid. You don't need a formula — you just need to be specific about what you need.

2. Iterate, don't perfect

I rarely get what I want on the first try. My typical workflow:

Write a decent prompt → get an okay response
"This part is too vague, add specific examples" → better
"Shorten the second paragraph and make the tone less formal" → good
"The third point is wrong, actually [correction]" → done

Each follow-up costs five seconds and dramatically improves the output. People who complain about ChatGPT's quality usually wrote one prompt and accepted the first response.

3. Break big tasks into small ones

Treating ChatGPT like a single-prompt machine is the fastest way to get mediocre output. Instead of "write me a complete marketing plan," try:

"Brainstorm 10 angles for marketing a developer tool"
"Pick the 3 best and write one-sentence value props for each"
"For angle #2, draft a rough content calendar for one month"
"Now write the first week's LinkedIn post"

Each step gives you a place to steer, correct, and improve. By the end, you have something that actually matches what you wanted.

Five Things I Actually Use ChatGPT For

Not hypothetical use cases. These are the things that show up in my chat history every week.

Debugging code I don't understand. Paste the error message, explain what you were trying to do, ask it to walk through the fix. I've saved hours on obscure TypeScript errors and dependency conflicts.

Rewriting things I've already written. I write a rough draft, paste it in, and say "make this clearer and cut 30%." The result keeps my voice but fixes my rambling. Way better than starting from a blank prompt.

Explaining concepts at different levels. "Explain database indexing to a junior developer," then "now explain it like I'm a CS grad student." The ability to dial complexity up and down is something textbooks can't do.

Generating boring boilerplate. CRUD endpoints, form validation, test fixtures — the stuff that needs to exist but doesn't require thinking. I write the interesting parts; ChatGPT fills in the scaffolding.

Playing devil's advocate. "I'm thinking of using MongoDB for this project. Argue against it." Forces me to confront trade-offs I'd otherwise ignore. I make better technical decisions because of these conversations.

Things That Will Waste Your Time

Expecting ChatGPT to know current events. Its knowledge has a cutoff date. For anything recent, use it alongside web search, not instead of it.
Asking for creative work with zero direction. "Write a story" gets you something bland and generic. "Write a story about a sysadmin who discovers the company servers have been running on a Raspberry Pi hidden in the ceiling" gets you something interesting.
Trusting math. Language models are bad at calculation. Use the data analysis tool (it can run Python) or verify separately.
One message = one task. If you're using ChatGPT like Google — one question, one answer, done — you're missing 80% of the value. The conversation is the feature.

What About Privacy?

OpenAI can use your conversations to improve their models unless you opt out. If you're pasting confidential code, business documents, or personal data — go to Settings → Data Controls and turn off "Improve the model for everyone."

For business use, look at ChatGPT Team ($25/user/month) or Enterprise plans. These don't train on your data.

I treat ChatGPT like email: fine for everyday stuff, but I wouldn't paste my tax returns into it.

Getting Past the Plateau

After a few weeks of using ChatGPT, you hit a point where it feels like you've seen everything it can do. You haven't. Here's what got me past that wall:

Custom GPTs: The GPT Store has specialized versions tuned for specific tasks. The "Data Analyst" GPT with code execution is genuinely more useful for spreadsheet work than the default chat. Take 10 minutes to browse what's available.
Memory: ChatGPT can remember things across conversations. Tell it your preferences, project context, or recurring constraints. Over time, responses get faster and more relevant because it already knows what you need.
Voice mode: On mobile, the voice conversation mode is surprisingly good for thinking through problems while walking or driving. I use it to brainstorm article outlines when I'm away from my desk.

FAQ

Is the free version good enough?

For most people, yes. GPT-4o mini handles everyday writing, basic coding, and general questions fine. Upgrade when you hit rate limits or need file uploads.

Can ChatGPT replace Google?

No. It doesn't have real-time information (without browsing mode) and it makes things up. Use ChatGPT for tasks that need synthesis, explanation, or generation. Use Google when you need current facts.

How is this different from Claude or Gemini?

ChatGPT is the best all-rounder. Claude writes better prose and handles longer documents. Gemini integrates with Google services. I use ChatGPT for most things, Claude for serious writing, and Gemini when I need YouTube or Maps data. More detail in my full comparison.

Will ChatGPT give everyone the same answer?

Not if you use Custom Instructions and provide context. Two people asking the same question can get very different responses based on their settings and conversation history.

Should I worry about ChatGPT taking my job?

Worry is the wrong word. Pay attention is the right one. The people I see thriving are the ones who learned to use AI tools to do more interesting work, not the ones who ignored them or the ones who let AI do everything. Tools change. Judgment doesn't.