A simple 5-step framework that turns mediocre ChatGPT answers into useful ones. With before/after examples for every step — so you can copy the lift, not just read about it.
Most beginner ChatGPT prompts fail in the same predictable way: they're too short, too vague, and assume the AI knows things it doesn't. The good news is that fixing this isn't a course or a certificate — it's five small habits, each of which takes seconds to apply. By the time you finish this guide, you'll be writing prompts that consistently come back useful on the first try.
ChatGPT, Claude, and Gemini are extremely capable, but they're also lazy in one important way: when you ask a vague question, they'll give you the safest, most generic answer that fits. That's not a bug — it's the right behavior given limited information. If you tell the model nothing about who you are or what good looks like, it'll guess. And the guess is almost always blander than what you needed.
The five-step framework below is just a way to give the AI enough information that it doesn't have to guess. Each step is simple. Stack them, and your prompts get dramatically better.
Tell the AI who to pretend to be. This is the single highest-leverage habit in prompting because the model has read millions of examples of how different professionals write — assigning a role unlocks a more specific, more useful voice.
Before: "Give me feedback on my resume."
After: "You are a senior recruiter who has hired hundreds of marketing managers at SaaS companies. Give me honest feedback on my resume below — what would you cut, what would you sharpen, and what's the single weakest line?"
Why it lifts: The first prompt gets you generic resume advice from a fictional career coach. The second gets you the kind of feedback an actual hiring manager would give — specific, opinionated, and useful. The role tells the AI what perspective to write from.
Tell the AI about your situation. Who's the audience? What industry? What constraint are you under? What have you already tried? You don't need a paragraph — two or three sentences usually does it. The goal is to give the model the variables it needs to produce a relevant answer instead of a generic one.
Before: "Write a cold email pitching my service."
After: "Write a cold email pitching my freelance copywriting service. I'm targeting early-stage SaaS founders who care about conversion on their landing pages. They've usually been burned by generic copywriters who don't understand product. They open emails on phones first thing in the morning."
Why it lifts: The second prompt gives the AI a specific buyer (early-stage SaaS founders), a specific objection ("burned by generic copywriters"), and a specific reading context (mornings, on a phone). All three of those reshape the email — shorter, more product-aware, less salesy.
State what you actually want. Use a single clear verb when you can. If your task has steps, number them. AI models do very well with explicit, ordered instructions.
Before: "Help me with my essay."
After: "Do three things with the essay below. (1) Point out the strongest paragraph and why. (2) Find the two weakest sentences and rewrite each one twice — once shorter, once with a clearer image. (3) Give me one structural suggestion for the overall essay. Keep your feedback honest, not flattering."
Why it lifts: "Help me with my essay" is so open the model has no idea what to do — it usually defaults to a polite, three-paragraph summary of vague suggestions. Numbered tasks force the AI to actually deliver each item separately, which makes the feedback far more useful.
Pin down what shape the answer should take. A bullet list of 7? A 200-word paragraph? A table with three columns? A numbered plan? This is the fastest way to stop AI from rambling — and it makes the output easier to use afterward, because you can drop a clean structure directly into Notion, an email, or a doc.
Before: "Compare iPhone and Pixel for me."
After: "Compare the iPhone 17 Pro and the Pixel 10 Pro in a 4-row table. Columns: feature, iPhone, Pixel, who wins. Rows: camera, battery, software updates, price. End with a one-paragraph recommendation for someone who values long battery life and Google services."
Why it lifts: The first prompt returns a wall of comparative paragraphs you have to skim. The second returns a table you can read in 30 seconds — plus a one-paragraph recommendation that does the synthesizing for you.
Tell the AI what not to do. Constraints are the secret weapon of good prompting because they let you remove the AI's blandest defaults. "Avoid jargon." "Under 100 words." "Don't use the word 'leverage.'" "No emojis." "Sound human, not corporate." Each constraint adds a small, useful filter to the output.
Before: "Write a LinkedIn post announcing my new role."
After: "Write a LinkedIn post announcing I just joined Acme as a senior product designer. 6 sentences max. Don't use 'thrilled to announce.' No emojis. Tone: warm, understated, no corporate-speak. End with a short, low-pressure invitation for product designers in fintech to message me."
Why it lifts: Without constraints, you'll get the LinkedIn post every LinkedIn post sounds like. With them, you get something that actually sounds like you — and the AI does the work of avoiding the clichés you don't want to read in your own post.
Here's a single prompt that uses all five steps. It's the kind of prompt that, once you write a couple of these, starts to feel natural in under a minute.
Full example: "You are a senior copy editor who specializes in product newsletters [role]. I'm writing a weekly newsletter for early-stage founders building B2B SaaS — they're time-poor and skeptical of fluff [context]. Edit the draft below for clarity and flow. Then suggest three sharper subject lines [task]. Show the edited version, then a numbered list of 3 subject lines [format]. Keep my voice — don't make it more formal. No clichés. Subject lines under 40 characters [constraints]."
That's a 75-word prompt. It hits all five elements without sounding stiff. The output you'll get will be miles ahead of "edit my newsletter."
Even with the framework, your first prompt rarely produces the perfect output. The fastest improvement most beginners can make isn't writing a longer prompt — it's getting comfortable replying to the AI to refine the answer. Three replies that work especially well:
If you remember just one thing from this guide: more specificity = better output. The five steps are just a checklist for being specific in the right places. You'll know you've internalized it when writing a good prompt feels less like work and more like writing a good email to a smart new hire.
Ready to apply this? Browse the beginner prompt library for ready-to-copy examples, or read ChatGPT vs Claude vs Gemini to pick the best tool for your kind of work.
Join readers learning to use AI better — one practical prompt and a short setup explaining why it works.
Unsubscribe anytime. We never sell your email.