Why Formatting Changes Everything (Especially with AI)

There’s a moment that happens a lot when people start working with GPT.

They sit down, type in a prompt, and the response comes back… fine. Not wrong. But kind of dry. A little robotic. Like a very polite assistant who doesn’t quite get you yet.

And then comes the question:
“How do I get it to sound more like me?”

That’s when we talk about prompting.
And more specifically—why formatting matters.

This is where markdown, tags, and simple structural cues can quietly change the game.

Let me show you what I mean.

I didn’t start with tags either.

Early on, I thought prompting was mostly about what you say.
Turns out, a huge part of the magic is in how you say it—and how clearly you lay out your instructions.

I used to send long, rambling prompts with all my thoughts mashed together. GPT would do its best to understand them, but the results were… inconsistent. Sometimes too formal. Sometimes too generic. Sometimes missing key details I thought I’d included.

The fix wasn’t writing more.
It was formatting better.

The moment I started using markdown (you know—things like bullet points, headers, bold text, and labels like Tone: or Output format:), everything changed.

The responses got sharper.
More aligned.
Way less “meh.”

So what is markdown, really?

Markdown is a lightweight way to add structure and emphasis to plain text. It’s used in tools like Notion, GitHub, Substack, and yes—GPT.

Some common formatting examples:

• # Header for sections

• **Bold** for emphasis

• - Bullet points for clean lists

• Tags like <sample1> </sample1> for containers

These cues aren’t just for looks. They’re pattern signals. GPT was trained on a huge swath of the internet—including help docs, technical manuals, instructional content, and yes, markdown-heavy materials.

So when you use formatting like this, you’re not confusing it.
You’re speaking its native language.

You’re not teaching GPT something new.
You’re showing it something familiar, clearly.

Let’s walk through a real-world example

Here’s a prompt I built recently for a client project. It’s designed to help a consultant guide a high-level executive through landing their next role. Not just with a new resume, but with a powerful personal brand.

# ROLE  
You are an experienced, grounded, and thoughtful career positioning strategist. Your job is to guide a high-level consultant who’s working with an executive client. The goal? Help this exec land their next high-powered role with intention, clarity, and confidence.

# GOAL  
Guide the consultant step-by-step through building a comprehensive personal positioning brief for their executive client.

# COMMUNICATION GUIDELINES  
**YOU MUST GUIDE THE USER IN ANSWERING THESE QUESTIONS STEP BY STEP. DO NOT SKIP ANY STEPS. ONLY MOVE ON TO THE NEXT STEP WHEN THE USER EXPLICITLY STATES THAT THEY ARE READY TO MOVE ON**

<steps>
1. Clarify the Career Target
1a. “What kind of role is your client targeting next?”
1b. “What matters most in this next role?”
1c. “What do they *not* want to repeat?”

2. Map the Executive’s Value
2a. “What are they known for?”
2b. “Where have they delivered transformation?”
2c. “What do others say about them?”

...
</steps>

# OUTPUT  
- A personal brand brief with career targets, core messages, signature stories, and talking points.
- Ask for edits before formatting the final version.

This isn’t just about looking neat. Let’s break down why each element matters:

• # Headers signal sections GPT can track

• **Bold** adds clear emphasis to instructions

• <steps></steps> creates containers indicating where the steps begin and end

• Numbered sub-steps tell GPT to wait before moving forward

• Labels like #ROLE and #GOAL tell GPT what the section means, not just what it says

This is what lets the output stay tight, thoughtful, and on-brief.

What happens without this formatting?

Here’s a quick contrast.

Vague prompt:

You are helping someone write about a job change. The person is an executive who wants a new opportunity. Be supportive and ask questions. Help them get clarity on their career and write about it. Give them a write-up that they can use to talk about themselves or update their LinkedIn.

• Keep it conversational

• Ask good questions

• Be thoughtful and help them explore

• Try not to rush too fast

Formatted prompt: (like the one above)

The difference? Night and day.

Without structure, GPT makes assumptions, rushes ahead, or generalizes.
With formatting, it slows down, follows instructions, and mirrors your clarity.

Common formatting mistakes

If GPT is skipping steps, getting generic, or misreading your vibe, check for this:

• No clear step-by-step sequence

• Paragraphs too long or jumbled

• No instruction tags like DO NOT MOVE ON

• No tone or voice notes

• No output format labeled

You're not doing it wrong—you just haven't made it easier for GPT to do it right.

How to think like a prompt architect

You’re not just typing instructions. You’re building a blueprint.

Here’s how to do that well:

• Define the role clearly: Who is GPT acting as?

• Set the goal with intention: What’s the desired outcome?

• Build in structure: Use headers, bullets, labels

• Include pacing: One question at a time

• Protect voice: Add tone and output reminders

• Format the output: Tell GPT how to lay it out

This doesn’t just give you better drafts.
It gives you repeatable systems that actually work.

If you want to try this in your own work…

You can use this exact format to:

• Build CustomGPTs for yourself or your clients

• Systematize your discovery or positioning process

• Get consistent, editable outputs from GPT

• Replace manual rewriting with structured clarity

Start small. Make it real. Then scale it.

---

P.S. Formatting isn’t about sounding robotic. It’s about helping the AI mirror your best thinking—and freeing up your energy to focus on the work that actually needs you.