A Guide to Prompt Writing
Why Your AI Prompts Matter More Than You Think
If you've ever asked an AI to help improve your writing and received something generic, robotic, or completely off-target, you're not alone. The difference between AI that feels like a helpful collaborator and AI that feels like a frustrating autocomplete comes down to the prompt.
AI reads your words and takes them very literally. A vague request gets a vague response. A specific, well-structured prompt gets something you can actually use.
Whether you're drafting emails, building apps, or crafting marketing copy, the quality of your prompts directly determines the quality of your output. This gets meta quickly because writing better prompts is a form of writing that AI can help you improve.
Is There Really Such a Thing as a "Bad" Prompt?
Absolutely. A bad prompt leaves too much to chance, forcing the AI to make assumptions about your audience, tone, structure, and goals. Sometimes the AI guesses correctly. Often, it doesn't.
A good prompt, by contrast, gives the AI the constraints and context it needs to produce something genuinely useful. Think of it like giving directions: "Go to the store" might eventually get someone there, but "Take a left on Oak Street, then your second right into the parking lot behind the blue building" gets them there faster and with fewer wrong turns.
Let's look at real examples to see this in action.
Example 1: Email Writing
The Bad Prompt
Write an email announcing that we are launching a new AI assistant tool called Acme AI.
What the AI Produced
Subject: Introducing Acme AI – Your New Intelligent Assistant
Dear Team,
We're excited to announce the launch of Acme AI, our new AI assistant tool designed to enhance productivity and streamline your daily workflows.
Acme AI is built to help you work smarter by providing intelligent support across a wide range of tasks. Whether you need assistance with research, content creation, data analysis, or problem-solving, Acme AI is ready to help you achieve more in less time.
Key Features:
- Natural language understanding for intuitive interactions
- Real-time assistance across multiple work domains
- Seamless integration with your existing tools and processes
- Continuous learning to better serve your needs
Getting Started: Acme AI is now available to all team members. To begin using it, simply [access instructions/link here]. We've also prepared comprehensive documentation and training resources to help you get the most out of this powerful new tool.
Why It's Bad
This email is fine—but that's the problem. It's generic corporate filler that could describe literally any AI tool. Notice the symptoms:
- Vague audience: "Dear Team" tells us nothing about who's actually reading this
- Generic benefits: "Work smarter," "enhance productivity," and "streamline workflows" are meaningless buzzwords
- No specifics: What does this tool actually do?
- Placeholder CTA: "[access instructions/link here]" shows the AI was guessing
- Forgettable subject line: Nothing compelling or specific
The Good Prompt
Your task is to write a compelling product launch email for "Acme AI,"
our new intelligent assistant integrated into the main platform.
Instructions:
- Target Audience: Existing enterprise users (Project Managers and Team
Leads) looking to reduce administrative overhead.
- Highlight 3 key capabilities: Automated meeting summaries/minutes,
"Smart Scheduling" (eliminating email back-and-forth), and predictive
task assignment.
- Tone: Innovative and empowering, but trustworthy. Avoid overused
buzzwords like "game-changer."
- Include a clear CTA: "Activate Acme AI Beta" (Directs users to their
settings panel).
- Subject line: Must include the 🤖 emoji and be under 45 characters.
- Personalization: Use {{First_Name}} to address the user directly.
Structure:
1. Subject Line
2. Email Body (Focus on "saving 5 hours a week")
3. Bulleted list of the 3 features
4. CTA Button Text
What the AI Produced
Subject: 🤖 Your AI teammate starts today
Hi {{First_Name}},
What would you do with an extra 5 hours every week?
That's the time Project Managers tell us they lose to scheduling meetings, writing up notes, and figuring out who should handle what task. We built Acme AI to give you those hours back.
Starting today, you have an intelligent assistant built directly into your platform—one that handles the administrative work that pulls you away from leading your team.
Here's what Acme AI does for you:
- Automated Meeting Summaries – Every meeting instantly becomes a clean, organized summary with key decisions and action items. No more frantic note-taking or "Wait, what did we decide?"
- Smart Scheduling – Tell Acme AI when you need to meet, and it finds the time that works for everyone. No more endless email chains trying to coordinate calendars.
- Predictive Task Assignment – Based on workload, expertise, and availability, Acme AI suggests the right person for each task. You make the final call, but the thinking is already done.
This isn't about replacing your judgment—it's about removing the friction that slows you down so you can focus on what actually matters: your team and your projects.
[Activate Acme AI Beta]
Why It's Good
The difference is night and day:
| Aspect | Bad Prompt Result | Good Prompt Result |
|---|---|---|
| Audience | Generic "Team" | Project Managers specifically |
| Benefits | Abstract buzzwords | Concrete: "5 hours a week" |
| Features | Vague capabilities | Three specific, named features |
| Tone | Corporate boilerplate | Human, empowering, specific |
| CTA | Placeholder text | Exact button copy provided |
| Subject | Forgettable | Emoji, under 45 chars, compelling |
The good prompt worked because it constrained the AI in all the right ways. It specified the audience, the tone, the structure, the exact features to highlight, and even formatting details like character limits.
Example 2: Vibe Coding
The same principles apply when you're using AI to write code—sometimes called "vibe coding." Let's look at a to-do list app with real-time sync.
The Bad Prompt
Create a to-do list app where I can add tasks with a time limit. Make sure
that if I have the app open in two tabs, the timer matches on both of them.
When I finish a task, just check it off but don't delete it. Save the data
so it doesn't get lost. Make sure everything syncs up properly.
Why It's Bad
This prompt has the right intentions but terrible specificity:
- "Time limit" – Input field? Dropdown? What format?
- "Timer matches" – What technology should handle this? BroadcastChannel? WebSockets? localStorage events?
- "Check it off" – Visual treatment? Strikethrough? Move to a different section?
- "Save the data" – LocalStorage? Database? Session storage?
- "Syncs up properly" – This means nothing without implementation details
The Good Prompt
Create a single-page to-do list application focused on real-time state
synchronization with the following specific requirements:
1. Task Creation Interface: A section with an input field for "Task Title,"
a numeric input for "Duration (minutes)," and an "Add Task" button.
2. Active Task List: Display tasks in a list format. Each row must include
the title, a live countdown timer, a "Start/Pause" button, and a
completion checkbox.
3. Cross-Tab Synchronization: Implement the BroadcastChannel API or
localStorage event listeners to ensure that creating a task, starting
a timer, or checking a box in one browser tab instantly updates the UI
in all other open tabs.
4. Live Timer Logic: When a timer is started, the countdown must visually
decrease in real-time across all synchronized tabs simultaneously.
5. Completion Behavior: When a task is marked as "Done" via the checkbox,
apply a strikethrough style to the text and stop the timer, but retain
the item in the visual list.
6. Data Persistence: Auto-save the entire list state (including active
timer values) to LocalStorage so data is restored upon refreshing
the page.
Why It's Good
This prompt succeeds by being:
- Architectural: Specifies BroadcastChannel API or localStorage events—actual implementation approaches
- Visual: Describes exactly what UI elements appear and where
- Behavioral: Defines what happens on each interaction (strikethrough, stop timer, retain in list)
- Sequential: Numbered requirements make nothing ambiguous
- Complete: Covers creation, display, sync, timing, completion, and persistence
Tips for Writing Better Prompts
Based on these examples and current best practices, here's what separates effective prompts from ineffective ones:
1. Be Clear, Contextual, and Specific
Don't make the AI guess. Provide:
- What the output will be used for (product launch email, internal memo, customer-facing docs)
- Who the audience is (Project Managers, developers, general public)
- What success looks like (specific metrics, tone, length, format)
2. Use XML Tags for Structure
When prompts get complex, XML tags help both you and the AI keep things organized:
<context>
You are writing for enterprise software users who are skeptical of AI hype.
</context>
<task>
Write three subject line variations for a product announcement.
</task>
<constraints>
- Under 50 characters each
- No exclamation points
- Include one emoji option and two text-only options
</constraints>
This approach offers:
- Clarity: Different sections are visually separated
- Accuracy: The AI can't misinterpret which part is which
- Flexibility: Easy to modify one section without rewriting everything
- Parseability: If you need the AI to output structured data, XML tags make extraction straightforward
3. Provide Instructions as Sequential Steps
Numbered lists force logical order:
1. First, analyze the current email for tone issues
2. Then, identify the three weakest sentences
3. Finally, rewrite only those sentences, keeping everything else intact
This prevents the AI from jumping ahead or missing steps.
4. Chain Your Prompts for Complex Tasks
For big projects, don't try to do everything in one prompt. Break it into a chain:
| Step | Prompt Focus | ||
|---|---|---|---|
| 1 | Analyze the source material and extract key themes | ||
| 2 | Create an outline based on those themes | ||
| 3 | Write the first draft following the outline | ||
| 4 | Edit for tone and clarity |
Chaining gives you:
- Accuracy: Each subtask gets the AI's full attention
- Clarity: Simpler subtasks mean clearer outputs
- Traceability: Easy to pinpoint where something went wrong
5. Iterate and Refine
Your first prompt is a draft, not a final product. If the output isn't right:
- Add constraints you forgot ("Make it shorter," "Avoid jargon")
- Provide examples of what you want
- Ask the AI to explain its reasoning, then correct course
Start With Prompts That Already Work
Writing great prompts is a skill—and like any skill, you don't have to start from scratch. Pre-built prompt libraries give you tested, optimized starting points that you can customize for your specific needs.
Explore Shaper's Prompt Library →
Our library includes prompts specifically designed to improve writing—including prompts that help you write better prompts. Yes, it's meta. But it works.
Whether you're drafting emails, improving documentation, or building AI-powered tools yourself, better prompts mean better results. Every time.
The quality of your AI output is directly proportional to the quality of your input. Start treating your prompts like the important writing they are.