Why do ChatGPT emails sound generic? A one-line prompt vs a real brief

One-line prompt vs brief, side by side

Read the two right columns side by side. Same task. Same tool. The only change is what you put in.

The one-line prompt has a faint role and nothing else. No goal, no context, no example, no don't-list. So the tool does the only thing it can. It writes the middle of the road, because most welcome emails are dull and you asked for one.

That is the point people miss. A bland draft is not a verdict on the tool. It is a verdict on the instruction. The tool was never the problem. The blank stare you handed it was.

Which two parts you are probably missing

Most failed drafts lack two of the four parts, and they are nearly always the same two.

The context. Tone is the big one. "Professional" means nothing to a machine. Write it how you would tell a new hire: "I write like a calm friend who happens to be good at this. No corporate filler." Add your limits too. Under 150 words. One clear next step. No more than one question. Limits are rails, not a cage.

The example. Almost nobody adds one, and it is the cheapest part, because you already own it. Find an email you were quietly proud of and paste it in. The tool reads its rhythm and its sign-off. One sample is enough to start. Two is better if they differ, like a warm one and a short one.

The don't-list is the part that grows. The first version is short. After a week you will have caught three new habits. Add them. A don't-list that never grows is frozen at the moment you knew least.

How to tell an email still sounds like AI

Before you send, read the draft once and look for the habits that give it away. These are the lines Nadia, the photographer in the book, put on her don't-list after she caught them more than once.

The book has a blunt test for the whole draft: if you could not tell who wrote it, the brief had no context. That one is not fixed by editing the email. It is fixed by adding the missing part to the brief, so the next draft comes back right.

Late at night, cold replies, lost couples

Nadia, the wedding photographer in the book, used to write every inquiry reply herself at 11 p.m. The late ones came out short. Some couples drifted to a photographer who had answered warmly by lunch.

She wrote one brief. The goal: send after a light edit. Her tone: warm, not gushy, no exclamation marks, no "your special day." Her three packages, and a rule to never quote a price in the first reply. One real reply she was proud of. A don't-list of every wedding cliché she had trained herself out of.

From the second inquiry on, the draft came back sounding like her. Her part shrank to a ninety-second edit.

I find this the most freeing part of the book. You were never bad at handing work off. You were handing it off without the year of context that makes it work.

Why this matters more when you run things alone

The best-known study of AI at work, by Brynjolfsson, Li and Raymond at the National Bureau of Economic Research, found the biggest gains went to the least experienced workers, up to 34% more per hour. That lift showed up when the worker was shown what good looks like. The brief is how you show it.

And the edge matters too. A Harvard and BCG study found AI users did worse than people without it on tasks just past the tool's range. A good brief keeps the tool on work it can do. A short check on prices and dates covers the rest. See what to check in AI drafts.

What The Leverage Stack adds to the brief

This is the problem The Leverage Stack was written for.

Chapter 3, The Brief, walks each of the four parts slowly, shows Nadia's full filled-in brief, and names the four ways briefs go wrong, from a role with no context to writing a brief for a task you barely do.

Chapter 4, The Pipeline, makes the brief attach itself every time, so it does not sit in a folder you never open. Read how to turn ChatGPT into a daily routine.

The Reusable Brief Template is in the free workbook.