The Best Prompting Advice Came From a Writing Teacher
The Best Prompting Advice Came From a Writing Teacher
William Zinsser spent his career teaching people to write clearly. His most famous book, On Writing Well, has sold over a million copies. And buried in it is a line that perfectly explains why so many people struggle with AI tools:
“Writers must therefore constantly ask: what am I trying to say? Surprisingly often they don’t know.”
That was written about writing. But it’s really about thinking. And thinking is the whole game when it comes to prompting AI.
The real problem isn’t the tool
When someone tells me they tried ChatGPT or Claude and “it didn’t give me what I wanted,” my first question is always the same: What exactly did you ask for?
Usually, the answer is vague. Something like “I asked it to write a letter to my client” or “I told it to help me with a contract provision.”
That’s the equivalent of walking into a restaurant and saying “bring me some food.” You’ll get something. But you probably won’t love it.
The AI isn’t failing. The person prompting it hasn’t done the hard part yet — figuring out what they actually want.
Zinsser’s insight applies directly
Zinsser understood that most bad writing isn’t a craft problem. It’s a thinking problem. People sit down to write before they’ve figured out what they’re trying to say. So they wander. They pad. They produce mush.
AI prompting works the same way. When you don’t know what you want, your prompt will be vague. And vague prompts produce vague results. Then you blame the tool.
But the tool was just following your lead. It gave you exactly what you asked for — which was nothing specific.
What “knowing what you want” actually looks like
This doesn’t mean you need a detailed outline before every prompt. It means you need to be honest with yourself about a few things:
What’s the purpose of this output? A letter to a nervous client reads differently than a letter to opposing counsel. If you don’t tell the AI which one you need, it can’t make that choice for you.
Who’s the audience? A memo for a senior partner and a summary for a client require completely different language. The AI can adjust — but only if you tell it to.
What’s the tone? Firm but fair? Friendly and reassuring? Blunt? You know what you need. The AI doesn’t.
What would make this a success? If you can’t describe what a good result looks like, you can’t evaluate what the AI gives you — and you definitely can’t guide it toward something better.
You don’t need to answer all of these in a formal way. But you need to have thought about them, at least roughly, before you hit enter.
A simple experiment
Next time you’re about to prompt an AI tool, pause for 10 seconds and ask yourself Zinsser’s question: What am I actually trying to say here?
If you can answer clearly in one sentence, you’re ready to prompt.
If you can’t — if your answer is something like “I just need it to, you know, help me with this thing” — you’re not ready yet. Spend another minute thinking. Write down the one sentence. Then prompt.
You’ll be surprised how much better the output gets when you do this. Not because you learned some advanced prompting trick. But because you got clear on what you wanted before you asked.
Bottom line
Most prompting advice focuses on technique — use this format, add these instructions, structure your prompt like this. That stuff has its place.
But the single biggest improvement most people can make is simpler than any technique. It’s the same advice Zinsser gave writers 50 years ago: figure out what you’re trying to say before you try to say it.
AI tools are powerful. But they can’t want something for you. That part is still your job.
;-)
Ernie
P.S. In the Inner Circle, we practice prompting together and learn from each other’s results — which is a lot faster than figuring it out alone. →
https://innercircle.ernietheattorney.net/