How to Write ChatGPT Prompts

Here's the single most useful mental shift for writing better prompts: stop treating ChatGPT like a search engine, and start treating it like a freelancer you just hired this morning.

Imagine you've brought on a smart, eager assistant for their very first day. They're capable, but they have three quirks. They can't see your screen, so they don't know what app you're in or what you've already drafted. They don't know your job, your colleagues, or your customers. And critically, they won't ask you follow-up questions unless you actively invite them to — they'll just guess and hand something in.

Once you internalize this, everything changes. You stop typing one-line questions and start writing short briefs. You include the background a real human would need. You say what "good" looks like. The rest of this guide is just practical ways to do that without it feeling like work.

Here's the formula that works for almost every everyday task. It has four pieces, and you can remember it as R-T-C-F:

Role — who should ChatGPT pretend to be? ("Act as a friendly career coach…") Task — what exactly do you want it to do? ("…rewrite my LinkedIn summary…") Context — what's the relevant background? ("…I'm a nurse switching into health-tech project work…") Format — how should the answer be shaped? ("…give me three short versions, each under 60 words.")

Stitched together it becomes one clear brief: "Act as a friendly career coach. Rewrite my LinkedIn summary. I'm a nurse switching into health-tech project work, and I want to sound approachable, not corporate. Give me three short versions, each under 60 words." That's a prompt that will get you something useful on the first try.

You don't have to use all four pieces every time, but if your answers feel generic, the missing piece is almost always one of these four.

The single most common beginner mistake is writing prompts that are too short. People worry that if they type too much, ChatGPT will get confused. The opposite is true — more relevant detail almost always produces a better answer.

Compare these two prompts with the same goal: plan a healthy week of dinners.

Short (5 words): "Healthy dinner ideas for the week." You'll get a generic list — quinoa bowls, grilled chicken, baked salmon — that could apply to anyone on the planet.

Detailed (about 50 words): "Suggest 5 healthy weeknight dinners for two adults. We're vegetarian, don't love mushrooms, and want each meal to take under 30 minutes. We have a stovetop and oven but no air fryer. Please list the meal, a one-line description, and the 5 main ingredients."

Same task. Completely different result. The second prompt gives you something you'd actually cook this week. Detail isn't overkill — it's the whole point.

"Add context" sounds vague, so let's make it concrete. There are four kinds of context that meaningfully improve almost every answer:

1. Who you are. Your role, your skill level, your audience. "I'm a small-business owner with no marketing background" tells ChatGPT to skip the jargon. 2. What you've already tried. If you've tested two subject lines and they flopped, say so. Otherwise it'll suggest those exact ones. 3. What you're optimizing for. Speed? Tone? Cost? Word count? Pick one and name it. 4. What to avoid. Negative constraints are surprisingly powerful — "don't use the word synergy" or "don't recommend anything over $50."

Quick example. Without context: "Help me write a cover letter." With context: "Help me write a cover letter for a junior data analyst role at a non-profit. I'm a recent grad with one internship. I want it to sound warm and curious, not stiff. Please don't use phrases like 'detail-oriented' or 'team player.'"

The second prompt does more than just sound better — it eliminates three rounds of editing.

If you don't tell ChatGPT what shape the answer should take, it'll default to wall-of-text paragraphs. That's fine sometimes, but often it's the reason a response feels useless even when the content is right.

Three formats cover most everyday situations. Use bullet points when you want to scan options quickly — brainstorming, comparing ideas, listing steps. Use paragraphs when you want flowing prose you'll send or publish — emails, messages, short articles. Use tables when you're comparing things across the same dimensions — pros and cons, three product options side by side, a study schedule.

Length matters just as much. Phrases like "keep it under 100 words," "give me three options," or "make this a one-paragraph reply" prevent the rambling answer that nobody reads. You can also ask for the structure explicitly: "Use the headings Problem, Why it matters, What to do." ChatGPT is great at following structural instructions when you give them.

One of the biggest unlocks for new users is realizing you don't have to nail your prompt on the first try. ChatGPT remembers everything in the current conversation. So if the first answer is 80% there, you don't rewrite from scratch — you just edit it like a draft with a collaborator.

Imagine you got a cover letter back and the tone is too formal. You don't need a new prompt. You can simply say: "Make it sound warmer and a bit less corporate. Keep the same structure." Or: "Shorten the second paragraph by half." Or: "Swap the example in paragraph three for something about teamwork instead of leadership."

This is how skilled users get great results. They don't write longer and longer prompts hoping to predict every requirement upfront. They start with a reasonable brief, see what comes back, and refine in 2–4 quick follow-ups. It's faster, less stressful, and the answers get sharper with each round — exactly like editing a Google Doc with a coworker.

You now have everything you need to get dramatically better answers from ChatGPT. Let's bring it all together.

The four-part formula: Role + Task + Context + Format. Treat ChatGPT like a smart freelancer on day one — explain what a thoughtful human would need to know. Write 2–4 sentences, not 5 words. Add the four kinds of context (who you are, what you've tried, what you want, what to avoid). Specify the shape of the answer. And refine with follow-ups instead of starting over.

A quick note on pitfalls: avoid asking yes/no questions when you actually want depth. Don't say "make it good" without defining what good means — say "warm and conversational" or "under 80 words" instead. And remember, ChatGPT isn't a search engine. It's a collaborator. The more you treat it that way, the more it gives back.

Your next step: pick one real task you'd normally do today — a tricky email, a meal plan, a topic you want explained simply, a draft you're stuck on — and write a Role + Task + Context + Format prompt for it. That single experiment will teach you more than any guide. You're ready.