AI keeps making the same mistake?
"Try again" was never going to stick.

The short answer

Two things are going on. A correction in one chat doesn't carry to the next unless it's written down somewhere your AI reads every time. And most corrections are too vague to learn from. "Try again" tells it something's wrong. It never says what right looks like.

Correct it like a new hire

When a new guy shows up on a job, you don't hand him a hammer and hope. A good new hire with no information does bad work. Your AI is that same new hire, every single day.

So you correct it the way you'd correct a person. Out loud. Specifically.

Bad correction vs. good correction

• Instead of "That's not right, try again," say "That doesn't sound like me. I'd never say 'excited to announce.' I'd say 'here's what I built.'"
• Instead of "Make it better," say "Too long. Cut it in half. Short sentences."
• Instead of rewriting it yourself and moving on, tell it exactly what was wrong so it stops doing it.
• Instead of fixing it in this chat only, write it down so it sticks.

The quick win: a corrections log

Open one note. Three columns: the date, what it got wrong, and what right looks like. Every time you fix something, add a line. For example: it wrote "We're excited to announce." Right looks like "Here's what I built." Direct, no hype words.

Hand that note to your AI at the start of a session. Every correction you write down is one you never make again.

One rule worth adding today

Give your AI this line word for word: "If you don't know something, say so. Never make up a fact, a number, or a result."

When it's bigger than a quick fix

It won't be perfect on day one. The first week, it'll still miss sometimes. That's normal. A corrections log works best when it sits next to the other things your AI should know every session: who you are, how you talk, what's already been decided, and where you left off. That's a setup, not a sticky note.

Why I'm telling you this

I spent 30 years building things with my hands. I had to close my company down and start over from zero. So at 53, off YouTube videos, I taught myself AI. I wasn't a tech guy. You don't need to be one either. For months my AI didn't sound anything like me. I'd explain my business on Monday, and on Tuesday it had no idea who I was.

Quick answers

Why does ChatGPT keep making the same mistake?
A correction in one chat doesn't carry to the next unless it's written down somewhere your AI reads every time. Vague corrections like "try again" also don't tell it what right looks like.

What does a good correction look like?
A specific one. Say what was wrong and show how you'd actually say it, like "Too long. Cut it in half. Short sentences."

How do I make a correction stick?
Log it. Keep a note with the date, what it got wrong, and what right looks like, and give it to your AI at the start of a session.

How do I stop AI from making things up?
Give it this rule word for word: "If you don't know something, say so. Never make up a fact, a number, or a result."

Will it be perfect right away?
No. The first week it'll still miss sometimes. What matters is that you correct it specifically and log the correction.

Educational only. Results vary and are not guaranteed. You're responsible for reviewing anything AI produces. Gen X Still Standing is independent and not affiliated with Anthropic, OpenAI, or Google. Built by a Gen Xer with AI, for Gen Xers.