Why most prompts fail, and the four-part fix

Most disappointing AI answers come from prompts that leave the AI guessing. "Write a post about my business" could mean a hundred things, so you get the most generic one.

A strong prompt answers four questions before the AI has to guess.

The four parts

1. Role: who should the AI act as? "You are an experienced copywriter for small Nigerian businesses."
2. Task: exactly what you want. "Write three Instagram captions announcing our new delivery service."
3. Context: what it can't know. "We're a bakery in Asaba. Delivery is free within 5 km. Our customers are mostly young professionals."
4. Format: what the answer should look like. "Each caption under 40 words, one emoji maximum, end with a call to order on WhatsApp."

Before and after

Weak:

Write a post about my bakery.

Strong:

You are an experienced copywriter for small Nigerian businesses.

Write three Instagram captions announcing our new delivery service.

Context: we're a bakery in Asaba. Delivery is free within 5 km.
Our customers are mostly young professionals ordering for the office.

Format: each caption under 40 words, at most one emoji,
and end with a call to order on WhatsApp.

Try it

Take a prompt you used recently that gave a so-so answer. Rewrite it with all four parts and run both. Keep the better version; it's the first entry in your prompt library.