The mistakes everyone makes

One message, expecting magic

You type one line, get a mediocre answer, and conclude "this isn't for me." But the first reply is a starting point, not the finished product. The real work happens in the back-and-forth.

Vague asks like "write me marketing"

The vaguer the request, the more generic the answer — not because it's dumb, but because you handed it almost nothing to work with.

Accepting draft #1

The first draft is rarely the best one — it's just the fastest one. Stopping there is like publishing your first rough draft because you were relieved to have anything.

Letting it flatten your voice

Left alone, ChatGPT writes in a smooth, generic, slightly corporate register — and if you publish that as-is, everything you make starts to sound like everyone else. Your voice is your biggest asset; don't hand it away.

The fix is simple: show it how you sound before asking it to write. Feed it three samples of your own writing first, then keep that voice guide handy.

Trusting the facts

ChatGPT states everything with the same confidence, whether it's certain or inventing. A made-up statistic looks identical to a real one. (Remember Lesson 1 — it predicts words, it doesn't look them up, unless it actually searches.)

Pasting in sensitive data

Customer social-security numbers, unreleased financials, someone else's private messages, passwords — what you type may be used to improve the models unless you've turned that off, and even then, treat the box like a postcard.

Using it to replace judgment, not multiply it

The goal was never to hand your thinking to a machine. When you outsource the decision instead of the drafting, the work stops sounding like you and starts sounding like nobody.