I had a wild thought the other day, and I can’t shake it: Your therapist asks better questions than you ask ChatGPT.
Think about it for a second. A good therapist dives deep. They ask things like, “What specifically made you feel that way? Can you give me an example? What would success actually look like to you in that situation?”
And what do we ask the most powerful information tool ever created? We type in: “Help with marketing.”
It’s almost comical. After watching over a thousand people try (and mostly fail) to get what they want from AI, I’ve seen the same mistakes over and over. I’ve started calling them the 7 deadly sins of bad prompting.
If you’re getting generic, useless, or just plain wrong answers from AI, I guarantee you’re committing at least one of these.
The 7 Deadly Sins of Bad Prompting
Let’s break down where things go wrong. These are the mistakes that turn a supercharged AI into a clunky, unhelpful intern.
-
SIN #1: 𝗡𝗢 𝗖𝗢𝗡𝗧𝗘𝗫𝗧
This is the biggest one. Asking AI to “Analyze this” is like handing a single page of a book to a friend and asking for a full review. It has no idea what matters!
❌ Bad Prompt: “Analyze this sales data.”
✅ Good Prompt: “You are a data analyst for a D2C shoe brand. Here is our sales data for Q3. Analyze it to identify our top-selling product, the time of day with the most purchases, and one surprising trend. I need to present this to my boss, who is non-technical.”
-
SIN #2: 𝗩𝗔𝗚𝗨𝗘 𝗚𝗢𝗔𝗟𝗦
AI isn’t a mind reader. “Write about marketing” could mean a million things. A tweet? A 10,000-word dissertation? A jingle? You have to define the finish line.
❌ Bad Prompt: “Write about marketing.”
✅ Good Prompt: “You’re a marketing expert specializing in SaaS startups. Your goal is to increase free-to-paid conversions. Create a 3-part email sequence for users on day 7 of their 14-day free trial. The tone should be casual, helpful, and create a sense of urgency.”
-
SIN #3: 𝗚𝗢𝗢𝗚𝗟𝗘 𝗠𝗢𝗗𝗘
You wouldn’t ask your star employee, “What is SEO?” You’d say, “Develop an SEO strategy for our new product line.” Stop asking your AI questions and start giving it tasks.
❌ Bad Prompt: “What are the best social media platforms for artists?”
✅ Good Prompt: “I am a painter who creates large, textured oil paintings of landscapes. My target audience is interior designers and art collectors aged 40-65. Create a social media plan outlining which 2 platforms I should focus on and why. For each platform, give me 5 content ideas.”
-
SIN #4: 𝗢𝗡𝗘 𝗔𝗡𝗗 𝗗𝗢𝗡𝗘
The single biggest lie about AI is that you get the perfect result on the first try. You don’t! The best results always come from conversation. The first output is just a starting point.
❌ Bad Prompt: Gets a mediocre response and gives up.
✅ Good Prompt: Gets a mediocre response and replies: “That’s a good start, but can you make the tone more professional? And replace the third point with something about ROI. Also, add some statistics to back up your claims.”
-
SIN #5: IGNORING THE PERSONA
Without a role, ChatGPT is just a generic language machine. Giving it a persona is like equipping a video game character with the right skills for the mission. It’s a total game-changer.
❌ Bad Prompt: “Write a post about leadership.”
✅ Good Prompt: “You are a former Navy SEAL who is now a leadership consultant for tech companies. Write a short blog post about how to lead a team through a high-pressure product launch.”
-
SIN #6: FORGETTING THE FORMAT
Don’t let the AI decide how to present the information. If you need a table, a list, or a JSON object, tell it! This saves you so much time on reformatting later.
❌ Bad Prompt: “Give me ideas for my podcast.”
✅ Good Prompt: “Brainstorm 10 podcast episode ideas for a show about vintage watches. Present them in a markdown table with three columns: ‘Episode Title,’ ‘Key Talking Points,’ and ‘Ideal Guest Type.'”
-
SIN #7: ACCEPTING THE FIRST ANSWER
This is about quality control. Sometimes the AI is lazy or hallucinates. Don’t be afraid to push back. Challenge it. Ask it to check its work or provide sources. You are the director, it is the actor.
❌ Bad Prompt: Sees a weird fact and just copies it.
✅ Good Prompt: “That’s an interesting statistic. Can you provide the source for that claim? Please double-check it.”
My Perfect Prompt Formula
So how do you avoid these sins? I’ve boiled it down to a simple, repeatable formula. It’s how your therapist gets great info out of you, and it’s how you’ll get great work out of your AI.
It goes like this: CONTEXT → TASK → FORMAT → EXAMPLES → ITERATE
- 💡 CONTEXT: Set the scene. Who is the AI? Who are you? What background information is essential?
- 🎯 TASK: Be specific. Use strong action verbs. What is the single most important thing you want it to do?
- 📝 FORMAT: Specify the output. A bulleted list? A table? A formal email? Code?
- ✨ EXAMPLES: Show, don’t just tell. Give it a small sample of what you consider a “good” output. “Here are 3 post headlines I love. Write 5 more in the same style.”
- 🚀 ITERATE: Treat it like a conversation. Refine, ask for changes, and guide it to the perfect final product.
The Real Problem Isn’t Artificial Intelligence
After all this, I’ve come to a conclusion. The problem isn’t the AI. It’s us.
For decades, we learned to communicate with computers using simple keywords and search queries. Now, for the first time, we have a machine that understands nuance, context, and goals. We just never learned how to ask.
Learning to prompt well is learning to communicate better. It’s a skill, and it’s one of the most valuable ones you can build right now. So, be the therapist. Ask the right questions.
Now, I want to hear from you. What’s the worst, most hilariously vague prompt you’ve ever sent to ChatGPT? Let’s have a laugh in the comments, we’ve all been there!