The Writing Prompts That Actually Work
Reviewed by Human · Updated August 25, 2026
You've been told a good writing prompt starts with *"Write a blog post about X."* That opening is the reason your drafts read like everyone else's drafts.
The model isn't a writer. It's a pattern-matcher running on the average of everything ever published about X. Ask for the average, get the average. The prompts that produce publishable work don't describe the *writing* — they encode the *decisions* a good writer makes before the first sentence.
✗ Wrong way
Write a blog post about remote work productivity.
Nothing here is a decision. No angle, no reader, no stance, nothing to leave out. The model fills every gap with the safest, most-repeated version of the topic — so you get an intro about 'today's fast-paced world', five obvious tips, and a conclusion that restates the tips.
✓ Better way
You're a remote-team lead writing for other team leads who already run distributed teams. Argue one contrarian point: that most 'productivity' advice quietly assumes an office. Open with a specific failure you'd have seen (a 9am standup that only works in one timezone). 700 words, no listicle, no 'in today's fast-paced world'. Exclude generic tips like 'use a good headset'.
Every added clause is a decision the model can't make for you: who it's for, what the single argument is, how to open, what to exclude. You did the thinking; the model did the typing. That division of labor is the whole game.
The fixed prompt wasn't better because it was longer. Something more specific changed. What?
Why Do AI Writing Prompts Produce Generic Content?
**Generic content**: output assembled from the statistical middle of a topic, where the model defaults when your prompt leaves the creative choices unmade. When you write *"Write about X,"* you've specified the subject and nothing else. The model resolves every remaining choice — tone, angle, structure, what to include — by reaching for the most common pattern in its training data. Common reads as bland. **Decision density**: the number of real editorial choices your prompt makes on the writer's behalf. Low density gives the model room to be average; high density boxes it into *your* voice and *your* argument. This is the same principle behind clear instructions beating vague ones, and it compounds with the context you give it — a sharper brief plus a sharper audience is multiplicative, not additive.
Your turn
Take a flat request and raise its decision density. Add exactly four decisions: an audience, an angle/stance, an opening move, and one thing to exclude.
Reflect
Which of your four decisions changed the output most? For most writers it's the angle, not the audience.
Two prompts request the same 800-word article. Which will most reliably produce non-generic writing?
You've finished this module.
Mark it complete to earn your XP and keep your streak alive.
Progress saved locally · Sign up to earn XP