The Prompt Spellbook: 12 Advanced Prompting Techniques for Methodical Magic

I read so many articles that basically amount to "say this magic spell" to prompt that I had to have fun with it—this article comes with a full 18 page Deep Research report on prompting techniques

My last post on prompting was super popular, but I realized I had left some advanced prompting techniques on the table. I built this post to highlight a dozen more advanced prompting techniques that are extremely high leverage—in other words these are techniques that pay to know and use. I organized them into pillars because I wanted to make it easier for you to learn, remember, and use them!

Large Language Models (LLMs) like o1 Pro, Deep Research, DeepSeek, o3-mini, and Gemini 2.0 Pro are powerful, but how you prompt them can dramatically change the quality of their outputs. While these models are deeply different in many ways, there are a few core techniques that are not commonly known or practiced that let you consistently upgrade the AI’s responses—whether you’re aiming for better logic, style, factual accuracy, or all of the above. And no, this is not about saying magic words. I don’t believe in spell-based prompting here, and I don’t think that saying words like “you are a magical SEO expert” meaningfully improves things. We don’t do magic here, or at least we are methodical about it. Instead I’ve written a guide for advanced prompting, structured under four strategic pillars:

1. Self-Review & Iteration: Getting the AI to refine its own output.

2. Multi-Stage & Collaboration: Chaining tasks and/or multiple models in a single workflow.

3. Logic & Precision: Ensuring step-by-step reasoning, structured formats, and minimal hallucinations.

4. Perspective & Style: Controlling tone, persona, or multi-voice simulations.

Within these pillars, we’ll cover 12 distinct techniques. Each technique has an introduction that lays out where and when to use it as well as the usual breakdown of What It Is & Why It Matters, Real-World Tie-Ins, Pitfalls & Solutions, and Extra Tips. And yes, we’ll link to a full 18 page discussion of these prompting techniques as prepared by Deep Research. Here we go!