Common Prompt Mistakes
The top errors people make and how to fix them.
Common Prompt Mistakes (And How to Fix Them)
Even experienced AI users make these errors regularly. The good news: once you learn to recognize them, they are easy to fix. Each mistake below includes a concrete example so you can spot the pattern in your own work.
Most bad AI outputs are not the AI's fault -- they are the result of a bad prompt. Fixing these eight common mistakes will immediately improve the quality of every AI interaction you have.
Mistake 1: Being Too Vague
Bad: "Help me with sales"
Fix: "Create a cold outreach email template for selling our B2B accounting software to CFOs at mid-size companies (100-500 employees)"
The vague version could produce anything from a sales training curriculum to a motivational speech. The specific version tells the AI the deliverable (email template), the product (B2B accounting software), the audience (CFOs), and the company size. No guesswork required.
Mistake 2: Asking Multiple Unrelated Things
Bad: "Write me a blog post about AI, also fix this Python code, and what's the weather?"
Fix: One prompt per task. Or explicitly separate them: "I have 3 unrelated requests. Please address each separately."
When you bundle unrelated tasks, each one gets less attention. The AI tries to split its focus and often does a mediocre job on all three instead of a great job on one.
Mistake 3: Not Specifying Format
Bad: "Give me marketing ideas"
Fix: "Give me 10 marketing ideas for a local bakery, formatted as a numbered list with a one-sentence explanation for each"
Adding format specifications is the easiest high-impact improvement you can make to any prompt. Try adding one of these to your next prompt: "Format as a table," "Use bullet points," "Number each item," "Keep each point to one sentence." You will be surprised how much more usable the output becomes.
Mistake 4: Forgetting the Audience
Bad: "Explain machine learning"
Fix: "Explain machine learning to a small business owner who has no technical background, using everyday analogies"
The same topic explained to a PhD student, a ten-year-old, and a CEO should look completely different. When you do not specify an audience, the AI defaults to a generic middle ground that is often too technical for beginners and too basic for experts.
Mistake 5: Not Iterating
Many people take the first output and give up if it is not perfect. This is like quitting a negotiation after the first offer.
First prompt: [Get initial output]
Follow-up: "Good start, but make it more concise and add specific dollar amounts"
Follow-up: "Perfect. Now adapt this for an email format instead of a report"
AI conversations are designed to be iterative. The first output is a draft, not a final product. Two or three follow-up prompts can take mediocre output and turn it into something genuinely excellent.
Mistake 6: Overloading Context
Do not paste 50 pages and say "summarize this." The AI will produce a generic summary that misses what you actually care about. Instead:
- Break large documents into sections
- Tell the AI what to focus on ("summarize the financial projections, ignore the company history")
- Ask specific questions rather than open-ended ones
Context overload also applies to system prompts. If you write a 2,000-word system prompt trying to cover every possible scenario, the AI will start ignoring or contradicting parts of it. Keep instructions focused and prioritized.
Mistake 7: Anthropomorphizing
Bad: "I hope you're having a good day! I was wondering if maybe you could possibly help me..."
Fix: Just state what you need directly.
You do not need to warm up the AI or soften your request. Small talk and hedging language add noise without improving output. Being direct is not rude -- it is efficient. That said, being respectful and clear is always a good practice.
Mistake 8: Not Verifying
Taking AI output at face value without checking facts, calculations, or code. AI models can sound extremely confident while stating something completely wrong.
Always verify anything you plan to act on -- especially numbers, dates, legal information, medical advice, and code that handles money or user data. The AI's confidence level has no correlation with its accuracy. A wrong answer stated confidently is more dangerous than an obviously uncertain one.
Exercises
0/3What is wrong with the prompt: "Help me with sales, fix my code, and plan my vacation"?
Find an AI conversation you've had before (or create one with a vague prompt). Identify which mistakes from this lesson were present, and rewrite the prompt to fix them. Test both versions.
Hint: Look for: vague language, missing format specs, no audience, no iteration.
Why is not iterating considered a mistake?