🔁 What do you do when the first answer misses?

Have you ever asked someone at the hardware store where something is, and they sent you to the wrong aisle? You don't give up and go home. You try again. "No, the little hooks. The ones for hanging pictures." The second try almost always works, because now they know what you meant.

FREE

If ChatGPT feels hard to talk to, my free Simple Prompts guide gives you ready-to-use lines you can copy, paste, and try today.

This is one of my free guides. There is nothing to buy, and it is yours to download and keep.

When you ask ChatGPT something, the first answer is a first try. Sometimes it's right on. Sometimes it's too long, too technical, or about something a little different from what you meant.

A lot of people take that as the end of it. They close the window and decide ChatGPT "isn't very good," or worse, that they aren't. But that first answer isn't a verdict. It's more like the clerk pointing you to the wrong aisle. You just need to say a bit more.

ChatGPT can't read your mind. It only knows the words you typed. If your question could mean two things, it picks one and runs with it.

Say you ask, "How do I make my phone easier to read?" It might tell you all about screen brightness, when what you meant was bigger text. Nothing's broken. It just guessed. One more line – "I mean the size of the words" – and you're there.

The AP put it nicely in a short guide to getting better answers: think of it like a conversation with a friend. You probably wouldn't end a chat after the first answer. Even OpenAI, the company that makes ChatGPT, says to adjust the wording, add more context, or simplify the request when the result isn't right. The AP's short guide is here.

That's freeing, if you think about it. The people who get good answers aren't smarter. They just keep talking.

When an answer misses, don't start over. Reply in the same chat, with one short line. Here are three that cover most situations:

1. Say what was wrong. "That's too long. Give me just the three main steps." Or, "I meant the phone, not the computer." Plain words work best.

2. Add the one thing you left out. "I'm using an iPhone." "This is for a friend's birthday." "I'm a beginner with this." One detail often changes the whole answer.

3. Ask it to ask you. "Before you answer again, ask me two questions about what I need." This one surprises people. It turns the guessing around, so ChatGPT finds out what you meant instead of assuming.

Here's a fourth, and it came from one of you. After the issue about asking ChatGPT to explain a word "like I'm a smart beginner," a fellow newsletter writer, Matthew Matchett, wrote back. He'd tried it with his own AI assistant, which he calls Val. Then he suggested a follow-up I hadn't thought of: "Now explain how I could use this in my own life or work." That's the step from understanding something to actually using it. I've been adding it ever since.

One honest caution. Asking again makes an answer clearer. It doesn't make it true. If the answer is about your health, your money or anything else that matters, check it somewhere you trust.

If you try one follow-up, you'll likely see the answer get better right in front of you. And you'll learn something useful: when ChatGPT misses, it isn't you. It's just a first draft.

If you don't, nothing's lost. You'll still get answers. You'll just settle for the first one more often than you need to.

The next time ChatGPT gives you an answer that isn't quite right, don't close it. Reply in the same chat with one line saying what was wrong, and see how the answer changes.

If you'd like, reply and tell me which of the three you tried first. Just the number is plenty, and there's no wrong answer.

Talk soon,
Bob

A short, no-pitch tip from my place – a two-minute way to quiet the clutter on your phone.

Nine plain questions about work you have already done, and my free tool hands back what you could teach and how a short guide on it would go.

This is my own free tool and it costs nothing to use. It carries on into my own course, where the first three modules are also free and the rest is $17 one time, so I do have a stake in you liking it.