How Self-Critique Prompts Sharpen Your AI Outputs

I want to share a technique that's quietly improved every AI output I've worked on, even when using detailed personas or structured frameworks like SCORE. It's simple: ask your AI to critique its own work.

Why This Works

Even if your setup is solid, you're still relying on a first draft. By asking the AI to evaluate its own response, you force a second pass, a chance to catch weak spots, rethink assumptions, and suggest sharper alternatives. It won't just refine your idea; it often points out flaws you didn’t notice.

My Process: Viking Content for TikTok

In my "Valhalla Awakens" project, I use ChatGPT to create Viking-themed motivational posts for my TikTok channel (@ValhallaAwakens3).

Here’s how it plays out:

Step 1: Set the Scene

ChatGPT takes on the role of a Viking sage. I open with "Brother," then feed it a quote like: *"Waiting for motivation is a trap. Discipline moves first."

Step 2: Get the Concepts

The AI gives me a breakdown of three content ideas, each with visual direction, rationale, and its preferred option. All solid, all thoughtful.

Step 3: Push It Further

Instead of picking one, I reply: *"Brother, can you critically critique each of the above concepts and see how you can improve them?"

Now it gets interesting.

The AI breaks down each concept:

• Strengths: What works well

• Weaknesses: What needs improvement

• Improvements: Specific changes to fix those weak spots

• Overall Verdict: A summary of which idea holds up best
  

And it does this all in Viking-speak.

This step consistently reveals insights I missed. Gaps in storytelling, lack of contrast, missed emotional beats, they surface here.

How You Can Use This

You don’t need a Viking project. This method works with any AI output. After the first response, prompt with:

• "Can you critically analyse each option and identify any weaknesses?"

• "What are the limitations of this response? How could it be improved?"

• "Review your answer and suggest refinements."
  

Repeat as needed. Each round tightens the outcome.

Works With Any AI Tool

This isn’t exclusive to ChatGPT. Claude, Gemini, or any LLM can handle this. The self-critique prompt simply triggers a different mode of thinking, one that's more reflective and analytical.

Bonus: Use ChatGPT Projects

If you're on the Pro plan, use Projects to keep related prompts together. It improves context and helps the AI deliver more consistent, relevant outputs across sessions.

Final Thought

Even if you're already good at prompting, there's always a better version one layer deeper. Ask your AI to critique its own work. The results speak for themselves.

Try it. You'll notice the difference right away.

See all my Digital Sparks here.