Most people's creative writing prompts look like this: "Write me a short story about a detective." The output is predictably bland.
The problem isn't the AI — it's the prompt. Here are techniques that consistently produce better results.
1. Give it a voice, not just a topic
Instead of: "Write a poem about loneliness." Try: "Write a poem about loneliness in the style of someone who is pretending not to be lonely — upbeat surface, undercurrent of ache."
2. Constrain to unlock
Paradoxically, more constraints produce more creative output. "Write a 6-word story about regret" forces the model to make every word count.
3. Seed with specificity
Vague prompts get vague output. The more specific the detail you provide — a particular place, a smell, a year, a mood — the more grounded and original the writing becomes.
4. Ask for the unexpected
"What would surprise me?" is genuinely useful. Adding "avoid clichés" or "don't use the obvious interpretation" pushes models past their default responses.
5. Iterate in conversation
The first output is a first draft. Respond to it: "The third paragraph is too safe — rewrite it to take more risk." AI creative writing works best as a dialogue, not a single request.