If your prompt is vague, your answer will be too. The fastest way to get a better answer isn't a longer prompt or a secret template — it's a more specific one.
The easiest way to learn what "specific" means is to see it. Below are five real prompts, exactly as someone might type them, next to a rewritten version. Same intent, same effort to read — but the second one tells the model who it's for, what the situation is, how to shape the answer, and what to avoid. That's the whole difference between a generic reply and one you can actually use.
The four things a vague prompt leaves out
Before the examples, here's the pattern every rewrite below follows. A vague prompt is usually missing some mix of:
- Audience — who the answer is for.
- Context — your situation, stack, or constraints.
- Format — how you want the answer shaped.
- What to avoid — the generic version you don't want.
Add those, and the model stops guessing at the safest middle and answers the question you actually meant. Now the examples.
1. Marketing — a LinkedIn post
Before:
write a linkedin post about leaving my senior PM role at a big tech company to start my own ai startup. want it to feel honest, not like every other "i quit my job" founder post out there.
After:
Write a LinkedIn post announcing my departure from a senior PM role at a major tech company to launch my own AI startup. The tone should feel authentic and genuinely reflective — avoiding the clichéd tropes of typical "I quit my job to become a founder" posts (e.g. calling it a "leap of faith," generic gratitude, or hype-focused language). Focus on honest insight into why this move made sense for me personally, what I actually learned, and what I'm genuinely excited about — not what sounds impressive.
What changed: the vague "not like every other post" became a concrete list of the tropes to avoid, plus a clear instruction on what to focus on instead.
2. Business — a pitch deck outline
Before:
draft a pitch deck outline for my ai startup. i'm raising a $2m seed round next quarter and want it to feel sharp, not the same generic 12-slide template every yc deck follows.
After:
Draft a pitch deck outline for a seed-stage AI startup raising $2M. Structure it to differentiate from standard narratives while maintaining investor appeal. Include: (1) a slide-by-slide breakdown with suggested content for each, (2) strategic positioning suggestions that differentiate from standard narratives, and (3) recommendations on which slides to emphasize vs. compress. Do not include design specs, color schemes, or implementation details — focus on narrative structure and content strategy only.
What changed: "feel sharp" became a numbered list of exactly what to produce, plus an explicit boundary on what to leave out.
3. Writing — a cold email
Before:
help me write a cold email to a podcast host i'd love to be on. don't want it to sound like every other "huge fan of your work" pitch sitting in their inbox.
After:
Write a cold email to a podcast host pitching myself as a guest. Make it stand out from generic fan pitches by being specific, authentic, and showing genuine familiarity with their work rather than using common boilerplate phrases like "huge fan of your work." I want it to feel personal and memorable.
What changed: the thing to avoid ("huge fan of your work") is named directly, and the goal ("personal and memorable") replaces a vague "don't want it to sound like…"
4. Coding — a React performance problem
Before:
fix performance in my react app. the lists are janky when scrolling — 500+ rows and it lags on every keystroke. don't just tell me to use react.memo, help me actually find the bottleneck.
After:
Help me fix performance lag in my React app where lists with 500+ rows cause stuttering during scroll and on every keystroke. I need you to help me identify the actual bottleneck causing the jank, not just suggest generic optimizations like React.memo. What diagnostic steps or profiling approach should I take, and what are the most common causes of this pattern?
What changed: the symptom is stated precisely, and the vague "help me find the bottleneck" became a real question — what diagnostic steps should I take — that the model can answer step by step.
5. Research — a concept explained
Before:
explain how transformers actually work, specifically the attention mechanism. i'm a cs undergrad new to ML, so go beyond the "queries, keys, values" soundbite without dumping a math textbook on me.
After:
Explain how the transformer attention mechanism works for someone new to ML. Start with the intuition behind why we need attention (the problem it solves), then build up to how queries, keys, and values actually work together. Use concrete examples or analogies rather than heavy math notation. Assume knowledge of basic linear algebra and neural networks, but not familiarity with attention or transformers.
What changed: the audience ("new to ML") and the format ("intuition first, then build up, use analogies") are spelled out, so the answer meets you where you are instead of guessing.
The pattern, in one line
Every rewrite above did the same thing: it named the audience, added the context, specified the format, and said what to avoid. None of them used prompt-engineering jargon. They're just more specific — and specific is what the model rewards.
You can do this by hand on every prompt. It works. It also takes a beat of effort each time, which is why most people skip it and then re-ask the same question five different ways.
If you'd rather not think about it, PrePrompt is a free Chrome extension that does exactly this rewrite in one click, right next to the Send button on ChatGPT, Claude, Gemini, and Perplexity — the five examples above are real PrePrompt rewrites. Type as usual, click Rewrite, send the sharper version.
TL;DR
- A generic answer almost always comes from a generic prompt.
- Good prompts add four things: audience, context, format, and what to avoid.
- You don't need jargon or length — just specifics. Often two or three extra sentences.
- The same pattern works across ChatGPT, Claude, Gemini, and Perplexity.
- Do it by hand, or let a one-click rewrite do it for you.