12 ways to make your prompts work better
(In case you had forgotten them)
There was a time when asking ChatGPT a question felt like magic. You typed a few words, hit enter, and watched the machine spit out ideas, answers, even whole business plans.
But now that the novelty has worn off, many people are hitting a plateau. The same vague answers. The same safe tone. The same feeling: “It’s fine, but it’s not quite what I wanted.”
If that sounds familiar, this post is for you.
Here are 12 simple, powerful prompting techniques to refresh your relationship with AI - and get better, faster, and more tailored results.
1. Be blunt about the format
Want a checklist? Ask for it. A table? Say so. A tweet thread? A limerick? A policy memo? Specify the format upfront. You’ll get something usable right away.
🧠 Try: “List the top 5 options as a pros/cons table.”
2. Use role prompting
Context is everything. Start your prompt with “You are a…” to shift the tone and focus.
🧠 Try: “You are an AI safety researcher. Write a risk assessment.”
3. Break it down (chain of thought)
Big question? Ask for step-by-step reasoning.
🧠 Try: “Talk me through this decision one step at a time.”
4. Give examples (few-shot prompting)
Don’t describe what you want - show it.
🧠 Try: “Here’s a good intro paragraph I like. Can you write three more in this style?”
5. Explain it like I’m five
Not just for kids. “Explain it like I’m five” is the fastest shortcut to clarity.
🧠 Try: “Explain GPT-4o to me like I’m five. Then like I’m a CTO.” But ALWAYS fact check please.
6. Ask for better prompts (meta-prompting)
Stuck? Ask ChatGPT how to improve your own question.
🧠 Try: “What’s a better way to phrase this prompt to get clearer results?”
7. Chain your prompts
Don’t expect perfection in one go. Use the output of Prompt 1 as the starting point for Prompt 2.
🧠 Try: “Take the last paragraph and rewrite it as a headline.”
8. Ask for context
If your question is too open-ended, turn the tables and let the AI ask you questions.
🧠 Try: “What else do you need to know before you answer this?”
9. Invite disagreement
Don’t let the model nod along to everything.
🧠 Try: “What’s wrong with this idea? Play devil’s advocate.”
10. Summarise in chunks
Too much to handle at once? Ask for section-by-section summaries.
🧠 Try: “Summarise this article paragraph by paragraph.”
11. Force ranking
Don’t settle for a list. Ask the model to prioritise, compare, or rank.
🧠 Try: “Rank these ideas by potential impact - and explain why you did that”
12. Timebox your prompt
Force the model to be concise.
🧠 Try: “Give me a 30-second elevator pitch. Then expand to 300 words.”
Your prompt is a prototype
Treat your prompt like a first draft. You’re not trying to be a prompt engineer—you’re trying to get what you need. If it’s not working, iterate. Ask questions. Push back. Think of the model as a slightly-too-eager intern: helpful, smart, but easily distracted. The more guidance you give, the better the work.
Keen to get better at everything AI? Join the waitlist for my Superhuman AI community. It’s only $25 a month and you’ll be joining me and a bunch of others on our journey to AI superhuman-dom.