Using AI to Write Better AI Prompts: A Meta Experiment

In 2023, I used ChatGPT to create the perfect prompt for… ChatGPT to rewrite articles. Then I asked ChatGPT to write a blog post about the process. This is that post, with some honest commentary added by an actual human.


The Experiment

I wanted to see if AI could optimize itself. Could I use an AI to create better instructions for that same AI?

Turns out: yes. Kind of. The AI helped refine a prompt for rewriting articles. The key improvements were:

  • Being explicit about what “rewrite” means (don’t just paraphrase – add depth)
  • Requesting specific formatting (headers, structure)
  • Asking for additional value (FAQs, examples)
  • Specifying tone while maintaining engagement

Did I do the hard work here? No. The AI did. I just asked the right questions and checked its output. (That parenthetical honesty was hand-written. The AI wanted to say “we collaborated.” We didn’t. It did the work.)


What I Learned About AI Prompts

Be specific. “Write a good article” gets generic results. “Rewrite this article to add depth, include practical examples, and maintain an engaging tone” gets better results.

Request structure. AI tends to ramble without guardrails. Ask for specific sections, word counts, or formats.

Iterate. The first output is rarely the best. Feed it back, ask for improvements, refine. This is where human judgment matters.

Add what AI can’t. AI doesn’t have your experience, your examples, your voice. The best results come from combining AI efficiency with human insight.


The Meta Question

Using AI to optimize AI prompts to generate AI content feels like we’re one step closer to The Matrix. I fed articles to a machine, it spit out improved versions, and now you’re reading content that’s… what, exactly? Part human, part machine, all weird.

The honest truth: AI makes certain tasks dramatically faster. But speed isn’t the same as quality. The articles AI rewrites are technically good. They’re also missing something – the specificity that comes from actual experience, the opinions that come from actually testing things.

We use AI as a tool now. We don’t publish AI output as-is. We’ve learned that lesson.


The Takeaway

AI prompt engineering is a real skill. Good prompts get dramatically better results than bad ones. But the output still needs human review, human editing, and human expertise to become genuinely valuable content.

If you’re using AI for content, don’t just accept what it gives you. Push back. Edit. Add your own experience. The AI is a starting point, not a finish line.

P.S. – This rewrite was done by a human (mostly). The 2023 original was mostly AI. Can you tell the difference? That’s probably the most interesting question of all.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.