AI can generate language at scale. It cannot decide what should be said.
Those decisions happen before prompts, and they’re the difference between content that converts and content that merely exists.
This is upstream thinking: the layer of judgment and taste that shapes everything downstream.
If upstream thinking is weak, AI makes the problem worse by producing more of the wrong thing, faster.
Good writing is a byproduct of good decisions
Most people think copywriting starts when the cursor blinks.
Nope.
It starts when someone decides:
what problem actually matters
what belief needs to change
and which angle is worth committing to
AI can mirror existing patterns but it can’t choose a point of view or recognise when the obvious angle is also the weakest one. It doesn’t feel when something is technically correct but strategically irrelevant.
Better output requires better decisions upstream.
Skill 1: Judgment
Judgment is the ability to say:
this matters, that doesn't
this sharpens the message, that dilutes it
this risk is worth taking, that one isn't
AI doesn't understand brand risk, audience fatigue, timing, or second-order effects. It has no idea when a clarification introduces doubt, or when an extra benefit weakens the main promise.
Humans understand because we're accountable for outcomes.
How can you sharpen your judgment?
Before drafting, answer three questions in plain language:
1. What belief needs to change for this to work?
If no belief changes, no action follows.
2. What is the strongest claim I can responsibly make?
Not the safest. The strongest you're willing to stand behind.
3. What would I cut if I had to remove 50% of this?
If you can't answer this, you don't yet know what matters.
After generating ideas (with or without AI), force yourself to:
rank them by impact
kill the safest option
commit to one primary message
If everything feels equally important, judgment hasn't been applied yet. If removing a paragraph doesn't change the outcome, it shouldn't be there.
Skill 2: Taste
What does your writing taste like?
Taste governs voice, restraint, and emphasis. It's knowing when to stop explaining, when to be blunt, or when a sentence is fine on paper but wrong in feeling.
AI smooths edges. It balances too carefully, explains too much, and resolves tension too neatly.
Human taste reintroduces sharpness, personality, and intentional omission to make writing feel authored instead of generated.
4 tips for developing your taste
Study great writing outside your niche
Rewrite strong copy from memory, then compare
Cut lines that "sound good" but don't earn their place
Prefer clarity first, then add edge deliberately
How to decide what NOT to say
Most weak copy fails not because it lacks ideas, but because it includes too many. AI amplifies this by producing exhaustive, evenly weighted output unless constrained by strong hierarchy.
1. Prioritise your messages
Take every potential point and categorize it:
Core: must be said for the message to work
Supporting: can be said if space allows
Distracting: feels useful but isn't
Delete the distracting points before you write.
This forces hierarchy instead of hoping it emerges later. It prevents AI from "helpfully" expanding minor ideas into full sections that dilute the main message.
If something isn't strong enough to be core, it doesn't deserve equal space.
2. Create early exit points
Good copy doesn't try to appeal to everyone. It repels as much as it attracts.
Decide in advance:
Where should an unqualified reader drop off?
Where should a qualified reader lean in?
These answers should shape your introductions, qualifiers, and transitions. They give the copy teeth and reduce the instinct to over-explain or hedge.
Is AI going to replace writers?
Writers who rely on AI to compensate for weak decisions will just get faster at being wrong.
But if you bring clarity, hierarchy, and restraint upstream, AI becomes a force multiplier instead of a liability.
So stop asking how to prompt better.
Start asking better questions. Decide what matters. Choose what to exclude. Commit to a point of view.
Because in a world where anyone can write with AI, the advantage belongs to the people who know what NOT to say—and are willing to stand behind the few things they do.
Like this article? Share it with a friend.
