How to Write Better AI Prompts: From Good to Great

The difference between a useful AI response and a useless one is almost always in the prompt. Writing better prompts is the fastest way to get dramatically more value from any AI tool. These techniques go beyond the basics — if you've already read the SUPER Formula lesson, this builds on it.

I'm Parikshet. I've been writing and testing AI prompts since I was nine. Here are the advanced techniques I use.

Chain of Thought Prompting

Adding "think through this step by step" to a complex question dramatically improves the quality of reasoning in AI responses. This works because forcing the AI to show its reasoning process catches more errors than allowing it to jump straight to a conclusion.

Compare: "What is 15% of 240?" (might just give you a number) vs "What is 15% of 240? Think through this step by step." (gives you the working and is less likely to be wrong).

Role Prompting

Starting a prompt with "Act as [expert]" changes the framing and often improves the quality for domain-specific questions. "Act as an experienced golf coach. I'm an 11-year-old with a handicap of around 18 who struggles with long putts. Give me 3 specific drills to improve my putting on courses with fast greens."

Role prompting isn't magic — the AI doesn't literally become an expert — but it calibrates the vocabulary, tone, and level of detail toward what that expert role would produce.

Output Format Specification

Being very specific about format produces dramatically better results. Instead of asking for "tips," ask for "a numbered list of exactly 5 tips, each under 30 words, written in simple English for a 10-year-old." Instead of asking for "a summary," ask for "a 3-sentence summary in paragraph form, then a 5-bullet-point key takeaways section."

Iterative Prompting

Treat AI as a conversation, not a single question. If the first response is almost right but misses something, tell it specifically what to fix: "This is good. Now make the language simpler — suitable for an 8-year-old. Also add one real-world example."

Each iteration narrows the gap between what you got and what you wanted. Professional prompt engineers sometimes do 5-10 iterations to get a final output exactly right.

Frequently Asked Questions

What is chain of thought prompting?

Adding "think through this step by step" to complex questions — forcing AI to show its reasoning, which reduces errors.

What is role prompting?

Starting with "Act as [expert]" to calibrate the AI's vocabulary, tone, and depth toward a specific domain.

The Iterate Rule (Most People Skip This)

The biggest secret to great AI results isn't the first prompt — it's the second, third, and fourth. AI rarely nails it on try one. Say "make it shorter," "add a real example," or "explain it for a 10-year-old." Treating AI like a conversation, not a vending machine, is what separates beginners from pros.

Try This

Take any boring prompt like "tell me about volcanoes." Now rewrite it using three upgrades: add a role ("you are a volcano scientist"), add a format ("give me 5 bullet points"), and add an audience ("for an 11-year-old"). Compare the two answers side by side — the difference will surprise you.