AI & Prompting

12 AI Prompting Tips: Stop Talking to AI Like a Bot

Most people use about 10% of what AI can do. These 12 tips cover a lot of the other 90%.

📅 December 2024 ⏱️ 10 min read ✍️ Diwakar Ray Yadav

Ask an AI for a "big creature in a desert" and you get a generic giant scorpion. Describe a dry world where moisture is scarce and people wear special suits to survive, and it comes back with sandworms.

That gap shows up everywhere. You've used ChatGPT or Claude. Sometimes the answer is great, but a lot of the time it's just mid: generic, robotic, or flat-out made up. Most of the time the fix is in what you type.

Treat AI like a basic Google search and you get basic results. Treat it like a creative partner and the answers get much better.

These 12 tips come from reading how experienced users prompt. They are practical things you can use today to write better essays, code faster, or explain ideas more clearly.

Part 1: Set up the context

The biggest mistake people make is lazy prompting: asking for a complex result without giving any context. The less effort you put in, the more average the answer.

1. The "world building" technique

AI defaults to the most average, generic answer possible because it doesn't know your context. Build the "world" for it first.

The idea:

Instead of only asking for an outcome, give the AI 3 to 4 distinct "puzzle pieces" of a scenario so it has to work inside a specific creative lane.

The "Dune" example:

If you ask AI for a "big creature in a desert," you'll get a generic giant scorpion.

But if you tell it: "Imagine a dry world made of sand. Moisture is scarce. People wear special suits to survive. What giant creatures live here?"

Because you built the world, the AI connects the dots and gives you something specific like "Sandworms." Give it the pieces and it will stitch them together.

2. The voice note context dump

Typing out a long backstory for world building is tedious. So don't type.

How to do it:

Open the mobile app, hit the voice-to-text button (not live mode), and ramble for 20 minutes. Tell the AI everything about your project, your constraints, and your goals. Then ask it to structure that messy dump into a usable context file. It's the fastest way to get what's in your head into the AI.

Part 2: Get specific answers

Vague prompts get abstract advice. These tips push the AI toward specifics.

3. Use few-shot examples instead of long instructions

Instead of writing long, complicated paragraphs telling the AI how to behave, show it.

The idea:

Structure your system prompts like simple code, with if/else logic and concrete examples.

Example prompt: "If the user asks for homework help, guide them step-by-step. If the user asks you to write the whole essay for them, politely decline. Example of a good response: [Insert Example]. Example of a bad response: [Insert Example]."

4. Meta-prompting: let the AI write the prompt

Not sure how to ask for what you want? Ask the AI to write the prompt for you.

How to do it:

If you want a very specific image from Midjourney or a complex web layout, tell ChatGPT your goal and ask it to generate the technical specifications.

Example prompt: "I want to build a landing page that looks like Stripe's. Break it down into components and give me the exact specs for the background gradients and colors so I can build it."

5. The multi-level explanation (study mode)

This one is for students. Ask for the same explanation at several levels of understanding.

Example prompt: "Explain [Quantum Physics] to me in three modes:
1. Like I'm a 5-year-old.
2. Like I'm 15 (plain English, no jargon).
3. Like I'm a 25-year-old physics major (include the math)."

Part 3: Keep the AI honest

AI likes to please you, so it sometimes says wrong things to make you happy. These prompts push back on that.

6. The "Red Pill" summary

Standard AI summaries are mostly facts you already know. This one asks for the ideas that go against common opinion.

How to do it:

When you upload a PDF or book, use this 3-step prompt:

"Summarize the main points."
"Give me the 'Red Pill' insights: the controversial or deep ideas in this book that most of the world doesn't believe."
"Give me actionable evidence for those insights."

7. Ask for confidence scores

To cut down on hallucinations, make the AI rate its own certainty.

Try this: Add this to your important prompts: "Answer only if you are confident. Provide a confidence score (0-100%) for every claim you make." If it gives a claim a 60% score, you know to double-check it.

8. The reality check (cost breakdown)

People throw around big numbers without understanding them. Use AI to check whether they make sense.

Example prompt: "People say a AAA game costs $100M to make. Break that down specifically. How much goes to developers vs. marketing? What are the specific salaries in that budget?"

Part 4: Use AI to improve yourself

AI can also act as a strict coach that points out where you are weak.

9. The gap finder

We don't know what we don't know. Use AI to find your blind spots.

Example prompt: "Based on the argument I just made (or everything you know about me), tell me: What are the gaps in my knowledge? Where is my reasoning flawed? Be brutally honest."

10. A personal learning feed

TikTok's algorithm feeds you entertainment. You can ask AI to feed you knowledge that fits your own interests.

Try this: "Based on my recent interests in [topic A] and [topic B], what niche concept should I learn about next that I probably haven't heard of?"

Bonus: two quick fixes

11. Remove the AI tells

AI writing has "tells": certain sentence structures that sound robotic.

The fix: Tell the prompt: "Avoid sentence structures that set up a statement and then negate it (e.g., 'X isn't just about Y, it's also about Z'). Use direct, affirmative sentences and vary your structure."

12. "Dark prompting" (emotional stakes)

Oddly, AI sometimes does better on complex tasks when you add emotional weight or "threats."

Try this: Adding phrases like "This is incredibly important for my career," or on the calmer side, "Take a deep breath and think step-by-step," can improve the quality of the output in complex math or logic tasks. Update: a 2025 Wharton study found that threats and tips like these no longer give a reliable boost on hard benchmarks, so treat this one as hit or miss.

What to take away

The people getting the most out of AI are the ones who communicate clearly with it, and that gap is growing fast. You don't need to know how to code to be on the right side of it.

Put real effort into your prompts. Build the world, ask for specifics, and use AI to find the gaps in your own thinking. Pick two of these tips and try them on your next real task.

About the author. Diwakar Ray Yadav writes about AI tools, prompt engineering, and automation from hands-on experiments. Read more about Diwakar.

Read next. If you want a practical list of current AI products, read Best Free AI Tools, LLMs, and Agent Platforms in 2026.

Also read. Prompt engineering changed shape this year. Read Prompt Engineering Is Dead. Here's What Replaced It.