Last updated on

How to Prompt ChatGPT: A Practical Framework

Two years ago I went deep on a research project where I tested seven generative AI tools to see whether they could replace human writers, especially for technical long-form content.

They failed. Hard.

But the technology moved quickly, and many of the limitations I hit back then have become smaller, or disappeared entirely.

Since then, I’ve spent a lot of time (both privately and at work) figuring out how to get consistently useful output from ChatGPT and similar tools.

The main lesson: good prompting isn’t magic. It’s structure.

A practical prompt framework

After a lot of trial and error, I landed on a simple framework that helps you get clearer, more accurate output—without turning every prompt into a novel.

The framework has five components:

  • Situation: Define the role of the AI
  • Precision: Specify the task clearly
  • Guidance: Set tone and constraints
  • Example: Provide a reference or pattern to follow
  • Output: Define the format you want back

It works because it forces you to provide the context you’d normally need when briefing a human.

Compare:

“Write a Facebook ad for my webshop.”

vs.

“You are a direct-response copywriter. Draft two Facebook ad variations for Danish parents buying children’s clothing. Start with a hook, use curiosity and urgency carefully, and output as two versions with headline + primary text.”

One gives you generic output. The other gives you something usable.

Example: using the framework

Here’s a concrete example:

Situation
You are a marketing copywriter for my webshop selling children’s clothing.

Precision
Draft an email newsletter that informs readers about a specific product/topic and gets them to click to read the full story.

Guidance
Open with a strong hook. Use psychological triggers like curiosity and fear of missing out—without being spammy.

Example
Here is a previous email that performed well. Follow the tone and structure.

Output
Write two variations:

  • one tailored to a Danish audience, primarily aimed at mothers
  • one aimed more broadly at fathers who could be in-market

The point isn’t “better prompts.” It’s better thinking.

These tools are getting better—but they still don’t replace judgment.

The win is combining AI speed with human intent and quality control:

  • use AI to accelerate drafts and exploration
  • use humans to decide what matters, what’s true, and what’s good

With a clear framework, prompting becomes repeatable—and you stop gambling on random outputs.

More writing connected to this topic, based on shared tags.