The Prompt Psychologist

The Prompt Iteration Workflow (Or: Why Your First Prompt Always Sucks)

7 min read

Here's a truth nobody wants to hear: your first prompt will be bad.

Not "needs a tweak" bad. Properly, embarrassingly bad. And that's fine.

Because prompting isn't about nailing it on the first try — it's about iteration.

Think of it like cooking. You don't make the perfect pasta on attempt one. You adjust salt, cooking time, sauce ratio. Same with prompts. Here's the workflow that actually works.

Step 1: The Vomit Draft Prompt

Start with whatever. Seriously. Just get something down.

Example first attempt: "Write a blog post about email marketing."

Output: Generic garbage about "building relationships" and "providing value." Useless.

But here's the thing: This uselessness is information. You now know what you DON'T want.

Too generic? Add specificity.
Too formal? Define tone.
Missing the point? Clarify goal.

The vomit draft isn't wasted effort. It's reconnaissance. You're testing the AI's defaults so you know what to override.

Step 2: Add Constraints Based on What Sucked

Look at the output. What's wrong with it?

Second attempt:

"Write a 300-word blog post about email marketing for solopreneurs. Focus on one counterintuitive tactic. Tone: conversational, not salesy. No fluff about 'building relationships.'"

Output: Better! More focused. But still a bit... bland.

You've fixed the structural problems (length, audience, focus) but it's still missing personality. That's normal. Structure first, personality second.

Step 3: Add Personality and Examples

Now that structure is solid, inject life into it.

Third attempt:

"Write a 300-word blog post about email marketing for solopreneurs. Focus on one counterintuitive tactic. Tone: like a friend who's slightly annoyed at bad advice and wants to set the record straight. Include one specific example. Avoid: corporate jargon, motivational fluff, anything that sounds like a LinkedIn post."

Output: Now we're cooking. Specific, opinionated, useful.

See what happened? You went from vague ("conversational") to specific ("friend who's annoyed at bad advice"). That's the difference between "meh" and "oh, this is good."

The Iteration Questions

After every AI response, ask yourself:

  1. Is this specific enough? Or is it vague platitudes?
  2. Is the tone right? Too formal? Too casual? Too boring?
  3. Does it have examples? Or just abstract concepts?
  4. Is it the right length? Rambling or too brief?
  5. Would I actually use this? Or does it need another pass?

If the answer to #5 is no, iterate again. Don't settle for "good enough" when "actually good" is one more prompt away.

Real Example: Iterating a Sales Email Prompt

Attempt 1 (Vomit Draft)

Prompt: "Write a sales email."

Result: Generic, robotic, instantly deletable.

What sucked: No context, no audience, no personality. The AI guessed "formal sales email" and gave you template #47.

Attempt 2 (Add Context)

Prompt: "Write a cold email to marketing directors at B2B SaaS companies. Selling: email deliverability tool. Length: 100 words."

Result: Better structure, still sounds like every other sales email.

What sucked: Has the bones, missing the personality. It's professional but forgettable.

Attempt 3 (Add Personality)

Prompt:

"Write a cold email to marketing directors at B2B SaaS companies. Selling: email deliverability tool. Tone: You're not selling, you're diagnosing a problem they might not know they have. Open with a surprising stat. Keep it under 100 words. No 'I hope this email finds you well' or any other tired openers."

Result: Actually interesting. Might get a reply.

What improved: Now it has a hook (surprising stat), a clear perspective (diagnosing, not selling), and specific things to avoid (tired openers).

Attempt 4 (Refine the Hook)

Prompt:

"Write a cold email to marketing directors at B2B SaaS companies. Selling: email deliverability tool. Open with this: '43% of your emails never reach the inbox. You just don't know which 43%.' Then explain the problem briefly, offer one quick win, and end with a low-pressure CTA. Under 100 words. No salesy language."

Result: Sharp. Punchy. Would actually work.

What changed: You gave it the exact hook, the structure (problem → quick win → CTA), and banned "salesy language." No room for the AI to guess wrong.

Why Iteration Beats Perfectionism

Trying to write the perfect prompt on attempt one is like trying to write a perfect essay in one draft. Doesn't happen. Even for people who do this for a living.

Iteration gives you:

The first draft is for discovery. The second is for structure. The third is for polish. Trying to do all three at once is why your prompts feel hard.

The "Save Good Prompts" Rule

Once you've iterated a prompt into something great, save it.

Don't reinvent the wheel every time. I have a doc of "prompts that actually work" that I reuse and tweak. Saves hours. Keeps quality consistent.

Here's what to save:

Build a library. Future you will thank you.

Common Iteration Mistakes

Mistake 1: Giving Up After Attempt 1

"AI is bad at this." No, your prompt was vague. Try again.

The AI isn't broken. Your instructions are just incomplete. One more iteration would've fixed it.

Mistake 2: Changing Too Much at Once

If you change 5 things between iterations, you won't know what worked.

Tweak one thing at a time. "Let me try adding a specific example" or "Let me define the tone better." Then you know what moved the needle.

Mistake 3: Not Diagnosing WHY It Failed

Don't just say "this sucks." Ask: Is it tone? Length? Specificity? Structure?

Fix that one thing. Then iterate again if needed.

Mistake 4: Accepting "Pretty Good"

If it's 80% there, one more iteration gets it to 95%. Don't settle.

The difference between "pretty good" and "actually great" is usually one prompt away. Push for great.

Advanced Move: The Refinement Loop

Once you have good output, you can make it great:

Follow-up prompt: "Now make this 20% more concise without losing key points."

Or: "Add one surprising detail that makes this more memorable."

Or: "Rewrite the opening to be more intriguing."

You're not starting over. You're refining what's already good. This is how you get from good to excellent.

When to Stop Iterating

You'll know. You'll read the output and think "yeah, I'd actually use this."

If you're on iteration 7 and still not happy, you probably need to rethink the entire approach, not just tweak the prompt.

But most of the time? 2-4 iterations gets you there.

The Mental Shift

Stop thinking of prompting as "telling the AI what to write."

Start thinking of it as "discovering what prompt produces what I want."

You're not writing instructions. You're running experiments. Each iteration is a test. Each result tells you something about what works.

Scientists don't nail the experiment on try one. Writers don't nail the essay on draft one. You won't nail the prompt on attempt one.

And that's fine. That's the process.

The Bottom Line

Prompting is editing. Your first draft is supposed to suck. The goal isn't perfection on attempt one — it's recognizing what needs fixing and iterating until it's right.

Embrace the mess. Iteration is where the magic happens.

Your first prompt is just a starting point. What matters is where you end up.