Most people using ChatGPT type one line, get something bland, and assume that is the ceiling. It is not. The difference between a generic answer and a genuinely useful one is almost always the prompt.
Below are 25 ChatGPT prompt examples grouped by job, written the way they actually work best: with a role, a goal and a format. Copy them, swap the bracketed parts for your own details, and keep the ones that earn their place.
The 5 parts every good prompt has
Almost every weak prompt is missing the same five things. Add them and the answer changes immediately.
- Role — who the AI should act as ("act as a direct-response copywriter").
- Goal — what the output is for, in one sentence.
- Context — who your customer is, what you sell, what you already tried.
- Format — bullets, table, 200 words, three options, plain English.
- Constraints — tone, reading level, what to avoid, what to never claim.
Writing and content prompts
- Act as an editor. Rewrite the text below so a busy reader understands it in one pass. Keep my meaning, cut filler, use short sentences: [paste text]
- Turn these rough notes into a clear 400-word article section with one subhead: [notes]
- Give me 10 headline options for [topic]. Five plain and specific, five curiosity-driven. No hype words.
- Summarize this document into five bullets a decision-maker can act on: [paste]
- Rewrite this paragraph at a 7th-grade reading level without sounding childish: [paste]
Marketing and sales prompts
- Act as a direct-response copywriter. Write three ad angles for [product] sold to [audience]. For each: the headline, the pain it names, and the promise.
- Write a landing page section that answers the objection "[objection]" in under 120 words.
- Draft a follow-up email to a lead who asked for pricing and went quiet. Friendly, one question, no pressure.
- Take this feature list and rewrite each line as a customer benefit: [features]
- Write five social captions for [topic] in my voice: [paste two samples of your writing]
Business and planning prompts
- Act as an operations consultant. Here is how [process] works today: [describe]. Show me the three steps most likely wasting time and what to do instead.
- Build me a simple weekly plan to launch [thing] in 30 days, with what to do each week.
- I have [budget] and [hours per week]. Rank these three options by realistic return: [options]
- Turn this messy meeting transcript into decisions, owners and next steps: [paste]
- Write a one-page brief for a freelancer doing [task], including what "done" looks like.
Customer and support prompts
- Write a reply to this unhappy customer. Acknowledge the problem, explain the fix, offer one concrete next step: [paste message]
- Turn our top 10 support questions into short FAQ answers, under 60 words each: [questions]
- Draft a refund policy explanation in plain English, no legal jargon.
- Rewrite this canned reply so it sounds like a person wrote it: [paste]
- Write three versions of a review request text message, each under 200 characters.
Personal productivity prompts
- Here is my to-do list. Group it into "today," "this week" and "delete," and tell me why: [list]
- Explain [complex topic] to me like I have 10 minutes and a real decision to make.
- Turn this goal into a checklist with the first step small enough to do in 15 minutes: [goal]
- Interview me with five questions to find out what is actually blocking [project].
- Read my draft and argue against it. What would a skeptical customer say?
Before and after
Before: "Write a social post about my coaching business."
After: "Act as a social media writer for a one-person business coach. Write a 90-word LinkedIn post for small business owners who feel busy but stuck. Open with a specific situation, not a question. End with one practical action. No emojis, no hashtags, no hype claims."
Same tool. Completely different output. The second prompt gave the model a role, an audience, a length, a structure and a set of constraints — which is all "prompt engineering" really is.
The fastest way to keep improving
Save the prompts that work in one place, and treat each answer as a first draft. When the output is close but wrong, do not start over: tell ChatGPT what to change. "Same structure, more specific examples, cut the intro" is often the single most valuable follow-up you can send.