I never thought my college coursework would become one of my most-used professional skills — not in SEO, not in marketing, and definitely not in AI. But here we are.
When I was studying for my Bachelor’s of Information Sciences at American Intercontinental University, one of the foundational courses was on pseudo code — the practice of writing out the logic of a program in plain, structured language before a single line of real code is written. At the time, it felt like busywork. Why describe what you’re going to code when you could just… code it?
Turns out, that “busywork” is exactly how I think through complex AI prompts today.
What Is Pseudo Code, Really?
Pseudo code is a way of expressing the logic of a program without worrying about syntax. It’s structured thinking in plain language. Something like:
IF user is a returning customer
THEN show personalized greeting
ELSE show standard welcome message
END IFThere’s no programming language here. No semicolons, no brackets. Just logic — clearly expressed, step by step.
The whole point of pseudo code is to force clarity before execution. You have to know what you want the program to do before you can write the code to do it. Vague thinking produces broken code. Clear thinking produces working code.
Sound familiar?
Prompts Are Just Pseudo Code for AI
When I started working seriously with AI tools — building workflows with Claude, GPT, and others — I noticed something immediately: the people struggling with AI weren’t struggling because the AI was bad. They were struggling because their thinking was vague.
“Write me a blog post about marketing” gets you a generic, forgettable output.
But if I think through the logic first — pseudo code style:
DEFINE audience: Marketing directors at B2B SaaS companies
DEFINE goal: Convince them to audit their attribution model
DEFINE tone: Direct, peer-to-peer, no fluff
DEFINE structure:
- Hook: A costly assumption most marketers make
- Problem: Why last-click attribution fails in long sales cycles
- Solution: Multi-touch attribution with specific examples
- CTA: Download our attribution audit template
OUTPUT: 900-word blog postThat prompt gets you something usable. Maybe even something great.
The difference isn’t the AI. It’s the thinking that went into the instruction.
The Pseudo Code Mindset Applied to Prompting
Here’s how I actually use this in practice. Before I write a complex prompt, I ask myself the same questions I’d ask before writing pseudo code:
- What are the inputs? — What context does the AI need? What data, examples, or constraints should it know about?
- What is the process? — What steps should it follow? Should it reason through something before answering? Should it consider multiple perspectives?
- What is the expected output? — What format? What length? What tone? Who’s going to read this?
- What are the edge cases? — What should it avoid? What assumptions might it make that would be wrong?
This is exactly the structure of a well-written program — and exactly the structure of a well-written prompt.
Chain-of-Thought Prompting = Pseudo Code in Action
One of the most powerful prompting techniques right now is chain-of-thought prompting — instructing the AI to reason step by step before giving a final answer. Researchers at Google found this dramatically improves accuracy on complex reasoning tasks.
But here’s the thing: chain-of-thought prompting is just pseudo code. You’re telling the AI to show its work — to walk through the logic before landing on an answer — the same way pseudo code walks through logic before producing executable code.
When I prompt for complex analysis, I’ll often include something like:
Before answering, think through:
1. What data points are most relevant here?
2. What assumptions might skew the analysis?
3. What would a contrarian argument look like?
THEN provide your final recommendation.That’s pseudo code. I’m defining the logic flow before the output. And the outputs are dramatically better for it.
What This Means If You’re Learning to Prompt
If you’re trying to get better at prompt engineering, here’s my advice: stop thinking about prompts as requests and start thinking about them as programs.
Every prompt has inputs, logic, and expected outputs. Every prompt needs to handle edge cases. Every prompt benefits from being written out in plain language before you try to “execute” it with an AI.
You don’t need a computer science degree to do this. You don’t even need to know how to code. But if you’ve ever written pseudo code — or even just thought through a problem step by step before acting — you already have the core skill.
The old skills don’t disappear. Sometimes they just need a new application.
Jonathan Alonso is a marketing professional specializing in SEO, AI-driven workflows, and digital strategy. He holds a Bachelor’s of Information Sciences from American Intercontinental University.
