Skill.reCommon Prompting Mistakes and How to Fix Them
Perfect prompts are rare. Good prompts are common. Bad prompts are everywhere. The difference between an okay prompt and a great prompt isn't talent—it's learning what mistakes to avoid. This lecture catalogs the ten most common prompting mistakes and shows you exactly how to fix each one.
As you read through these, you'll probably recognize yourself making several of them. That's normal. Everyone does. The point is to become aware so you can fix them. Each mistake comes with a before-and-after example so you can see the exact difference a fix makes.
Mistake 1: Too Vague
The Problem:
Your prompt lacks specificity. "Write an email" or "Give me ideas" doesn't tell the AI anything about your situation, who the audience is, what you're trying to achieve, or what good looks like.
The Fix:
Add context about who, what, why, and success criteria.
Why this matters: Vagueness forces the AI to guess. Specificity lets it deliver exactly what you need. This single mistake causes most of the disappointing AI outputs people get.
Mistake 2: Too Long and Unfocused
The Problem:
Your prompt is a rambling, disorganized brain dump. Multiple tasks, unclear priorities, conflicting requests. The AI doesn't know what's actually important.
The Fix:
Focus on one thing at a time. Be organized. Save other topics for separate prompts.
Why this matters: Focused prompts produce focused outputs. Long, unfocused prompts produce long, unfocused, mediocre outputs. If you have multiple topics, write multiple prompts.
Mistake 3: Missing Critical Context
The Problem:
You assume the AI knows things it doesn't know about your business, your market, your constraints, or your situation. Generic advice results because the AI lacks the context to be specific.
The Fix:
Provide context about your actual situation, market, constraints, and current performance.
Why this matters: Context turns generic advice into tailored recommendations. A solution that works for a 50-person company might be wrong for a 5-person company. Tell the AI your reality.
Mistake 4: Not Specifying Format
The Problem:
You don't tell the AI how you want the output structured. The AI picks its own format. You spend 15 minutes reformatting to actually use it.
The Fix:
Specify exactly how you want the output formatted so it's immediately usable.
Why this matters: Format specification is a force multiplier. It takes 30 seconds to specify, saves 15 minutes of editing, and makes output immediately useful.
Mistake 5: Not Iterating
The Problem:
You get output, you use it as-is. You don't ask follow-ups, request refinements, or iterate toward better results. You accept the first draft.
User: "Write a job description"
AI: [Returns job description]
User: Done, posting this.
The Fix:
Iterate. Ask follow-up questions and refinements until it's right.
User: "Write a job description"
AI: [Returns job description]
User: "This is generic. Add more about the team environment and emphasize we're a startup. Make it more personality-driven."
AI: [Refined version]
User: "Better. Remove the 5 years experience requirement—we'll train the right person. Add a line about our commitment to work-life balance."
AI: [Final refined version]
User: Done, this is perfect.
Why this matters: Great outputs usually require 2-3 iterations. The first draft is rarely perfect. Treat prompting as iterative refinement, not one-shot creation.
Mistake 6: Treating AI Like a Search Engine
The Problem:
You ask generic questions like "What are the top 10 ways to reduce support costs?" without providing your specific situation. You get generic lists.
The Fix:
Provide your specific situation so you get tailored advice, not generic lists.
Why this matters: AI is not a search engine. You're not looking for "what does the internet say." You're getting advice from an expert consultant. That consultant needs your specifics to give useful advice.
Mistake 7: Not Specifying Audience or Level
The Problem:
You don't tell the AI who will read this or what their knowledge level is. The AI guesses. Output might be too technical or too basic, the wrong tone, or inappropriate for your audience.
The Fix:
Specify who the audience is and what they understand.
Why this matters: Audience matters tremendously. Explaining data retention to engineers is completely different than explaining it to non-technical customers. Specify audience so the AI writes for the right people.
Mistake 8: Not Assigning a Role
The Problem:
You don't tell the AI what perspective to take. "Write a recommendation" is vague about how the recommendation should be framed.
The Fix:
Assign a role that shapes the perspective and reasoning.
Why this matters: A strategist thinks about direction. An engineer thinks about technical debt. A CFO thinks about costs. The role you assign shapes the entire output. Choose roles strategically.
Mistake 9: Trying to Do Everything in One Prompt
The Problem:
You ask the AI to analyze data AND identify problems AND recommend solutions all in one prompt. Complex tasks get shallow treatment.
The Fix:
Break complex work into sequential steps. Do each one thoroughly.
Step 1 (Analyze): "Analyze this customer feedback from our exit interviews: [paste feedback]. Identify the top churn reasons with frequency."
Step 2 (Diagnose): "Based on these identified churn reasons: [paste AI's analysis]. For each reason, diagnose: Is this a product issue, support issue, pricing issue, or usage issue?"
Step 3 (Recommend): "Based on this diagnosis: [paste previous output]. For each churn reason, recommend specific retention strategies we could implement in 30 days."
Why this matters: Multi-step prompting produces deeper analysis. Each step gives the AI room to think thoroughly about one thing. This is better than trying to squeeze everything into one prompt.
Mistake 10: Not Providing Examples
The Problem:
You describe what you want but don't show an example. The AI guesses at your tone, style, structure, or approach. Output misses what you actually wanted.
The Fix:
Provide an example of your desired tone, style, or structure.
Why this matters: Examples are worth a thousand words. If you have existing content that shows your style, paste it. The AI will match the tone and style perfectly instead of guessing.
Self-Assessment Checklist
Before you hit send on any important prompt, run through this checklist:
Prompt Quality Checklist
- Is this specific enough, or could it apply to anyone?
- Did I provide enough context about my situation?
- Is the task or request clear and focused (one main thing)?
- Did I specify the format I want the output in?
- Did I mention who the audience is or what level they're at?
- Did I assign a role if the perspective matters?
- Did I add length constraints or other limiting parameters?
- Did I provide any examples of tone or style if that matters?
- Is this the most important thing to ask, or should I break it into steps?
- Am I planning to iterate, or just use the first output?
You don't need a perfect check on all 10 items. But if you're checking fewer than 6, your prompt is probably too vague and will produce disappointing output.
Key Takeaway
The ten most common prompting mistakes are: being too vague, being too long and unfocused, missing context, not specifying format, not iterating, treating AI like a search engine, not specifying audience, not assigning a role, trying to do everything at once, and not providing examples. Each one is easy to fix once you know about it. Use the self-assessment checklist before important prompts. Start recognizing these mistakes in your own work and fixing them. This is how you move from average prompts to great prompts that consistently deliver useful output.
What You'll Learn Next
Now that you know how to avoid mistakes and write better prompts, the final lecture shows you how to build a personal library of prompts you reuse. teaches you systems for organizing, managing, and continuously improving prompts for all your recurring business tasks.
Frequently Asked Questions
What is the most common prompting mistake?
The most common mistake is being too vague. Prompts like "write an email" or "give me ideas" lack context and specificity. The AI can't know what you actually need, so it produces generic output. Always include who, what, why, and what success looks like. The difference between vague and specific prompts is often the difference between useless and incredibly useful output.
How long should a prompt be?
There's no magic length. A prompt should be as long as it needs to be to provide necessary context and clarity. A one-sentence prompt might be perfect for a simple task. A three-paragraph prompt might be necessary for complex strategic work. Length matters less than specificity and clarity. Focus on whether the AI has what it needs to deliver what you want.
What does "not iterating" mean as a prompting mistake?
Not iterating means accepting the first output without asking follow-up questions or requesting refinements. Great output usually requires multiple turns: initial response, then "more detailed," then "adjust the tone," then "add X element." Prompting is iterative. Refine until you get what you need. Many people treat AI like it gives one-shot answers, but the best results come from treating it like a conversation.
Why is treating AI as a search engine a mistake?
AI tools aren't search engines—they're thinking partners. Using them like search engines ("What are the top 10 ways to reduce costs?") produces generic lists. Treating them as consultants ("Here's our situation... what's the best way for US to reduce costs?") produces tailored advice. The difference is context and specificity. Generic questions get generic answers. Specific situations get specific recommendations.
What's the difference between one-shot and multi-step prompting?
One-shot prompting asks the AI to do everything at once: "Analyze this data, identify issues, and recommend solutions." Multi-step prompting breaks it into stages: First analyze, then identify the top 3 issues, then recommend solutions for each. Multi-step often produces better results for complex work because each step gets full attention rather than being squeezed into one response.