What makes a good prompt vs a bad prompt? I can show you faster than I can tell you. Here are two real prompts I used last month for the same task — writing a product launch email:
Bad prompt: “Write an email about our new feature.”
Good prompt: “Write a launch email for our new time-tracking feature. Audience: freelance designers who bill hourly. Goal: get them to try it free for 14 days. Tone: friendly, direct, no corporate jargon. Structure: hook (1 sentence), the problem (2 sentences), the feature (3 bullet points), CTA. Keep it under 180 words. Here’s an example of our voice: [pasted previous email].”
The first got me generic mush. The second got me a draft I sent with light edits. Same AI, same task — wildly different results. Let’s break down exactly why.
The 5 Differences Between Good and Bad Prompts
1. Context: Good Prompts Brief the AI
Bad prompts assume the AI knows your situation. It doesn’t. Good prompts include: who this is for, what the goal is, and any background that matters. Think of briefing a smart freelancer — you’d never say “write an email about our feature” to a human either.
2. Specificity: Good Prompts Say Exactly What “Good” Looks Like
“Write a summary” is a bad prompt. “Summarize this in 5 bullet points, each under 20 words, focusing on action items” is a good prompt. Vague prompts get vague outputs — the AI fills the gaps with guesses, and its guesses are generic by design.
3. Format: Good Prompts Specify the Shape of the Answer
Want a table? Say so. Want bullet points? Say so. Want exactly 200 words? Say so. The AI can’t read your mind about format, and “whatever looks good” produces whatever’s average. Our prompt structure guide has a reusable template for this.
4. Examples: Good Prompts Show, Don’t Just Tell
This is the highest-leverage difference. Pasting one example of what you want (“here’s the tone/length/format I’m after”) improves output more than any other single technique. Bad prompts describe; good prompts demonstrate.
5. Constraints: Good Prompts Say What to Avoid
“Don’t use buzzwords.” “No exclamation marks.” “Don’t mention pricing.” Constraints are criminally underused. The AI’s default is generic-marketing-voice; constraints carve away the generic until only the good stuff remains.
Side-by-Side: More Examples
Task: summarize a meeting transcript
- Bad: “Summarize this meeting.”
- Good: “Summarize this meeting transcript. I need: 3 key decisions (with who made them), all action items (with owners and deadlines mentioned), and any unresolved disagreements. Format as bullet points under those three headings.”
Task: explain a concept
- Bad: “Explain blockchain.”
- Good: “Explain blockchain to a smart 12-year-old. Use one everyday analogy, keep it under 200 words, and end with one sentence on why people care about it.”
Task: analyze data
- Bad: “What do you think of these sales numbers?”
- Good: “Here are our Q3 sales by region [data]. Identify: the top 3 trends, any anomalies vs Q2, and one hypothesis for why the West region declined. Present as a brief with headings.”
Why Bad Prompts Feel Like the AI Is Dumb
Here’s the uncomfortable truth: when the AI gives you a bad answer, it’s usually your prompt. The model is doing its best with insufficient information — and “its best” with no direction is aggressively average. I’ve tested this myself hundreds of times: the same model that writes mush from a lazy prompt writes brilliantly from a good one.
Before blaming the AI, try the upgrade test: add context, specificity, format, an example, and constraints. If the output is still bad, then it’s the model. Nine times out of ten, the rewrite fixes it.
Level Up From Here
- Learn the reusable structure: how to structure prompts for best results
- Stop guessing which version works: how to test prompt versions
- The bigger picture: is prompt engineering still relevant in 2026
What Makes a Good Prompt vs a Bad Prompt in Real Examples?
Let me make this concrete with a before-and-after. Bad: “Write about marketing.” Good: “Write a 300-word LinkedIn post for B2B SaaS founders about email deliverability. Tone: direct, no hype. Structure: open with a surprising stat, then three actionable tips, then a question. Avoid jargon like ‘synergy’ and ‘leverage.'”
See the difference? The good prompt specifies audience, length, format, tone, structure, and constraints. Each element removes a guess the AI would otherwise make randomly. You’re not being fussy — you’re being a good manager giving a clear brief.
The test I use: could a competent human freelancer produce what I want from this prompt alone? If yes, the AI probably can too. If the prompt only makes sense with context in my head, it needs more detail. For the underlying principles, the documented prompt engineering techniques formalize what experienced users learn by trial and error.
The Bottom Line
What makes a good prompt vs a bad prompt? Context, specificity, format, examples, and constraints. Bad prompts are vague wishes; good prompts are clear briefs. Master these five elements and you’ll get dramatically better results from every AI you use — not through tricks, but through the simple discipline of saying what you actually want.
