You type a request into ChatGPT and take what comes back. It fills the gaps with assumptions about what you meant, and those assumptions change the output.

When ChatGPT goes off in the wrong direction and you don't catch it early, you could spend another hour refining until you get the right thing. Five minutes at a time, one small change at a time, when you could have just got it right from the start.

An hour a day of this becomes five hours a week. Multiply that by your hourly rate, and you're talking thousands. These four words are your get-out clause. Think of this as sharpening your knife before you make the cut.

Why ChatGPT goes in the wrong direction

Whenever people normally use ChatGPT or Claude, they ask it for something and they assume that the LLM will fetch it for them. But the LLM needs more context. Often, if you don't give it permission to ask you questions, it won't. It will barrel forward with its first idea, and it might go in a wildly different direction to the one you had in mind.

When that happens, you're in the business of re-prompting. You send clarification. It makes the edits, but it's still going off its original idea, which might be far from what you wanted.

The tool will keep making its best guess, trawling through the chat history to figure out what you mean, but things change. You change your mind. You’ve been influenced by what you’ve seen. The goal posts for this piece of work have moved, and ChatGPT doesn't know that.

The four words that fix it

So the four words are this:

“Ask me questions first.”

When you use these four words, the AI tool you're using knows that it needs to scrutinise. It knows that there's more to this than your original prompt. It's likely to ask you a lot of questions, filling in the gaps you missed.

These questions might be about the direction of the work. They might be about the purpose of the work. Perhaps something you said in your original prompt didn't quite stack up, and it wants to double-check. Either way, you prevent expensive and time-consuming mistakes with these four simple words.

What it looks like in practice

Imagine you ask ChatGPT to write a LinkedIn post. You give it all the context. You give it your past posts. You give it what you want to talk about and details about your dream client. If you don't say "ask me questions first," it'll produce a post that seems passible, but on closer inspection, it's hollow, inauthentic, and reads like it was written by a robot.

Now you have a decision to make. Do you re-prompt, make edits, and ask for it to be different? Do you edit it yourself, removing the whole point of the exercise? You might have written something better from scratch. You don't want this decision. You've got a business to run.

If you say "ask me questions first," it's going to dig deeper into your beliefs. It's going to understand more of what you're about and what you want, and when you get the final output, it's going to be something you’re proud to post.

Beyond content creation and planning, this question will improve your search results. Instead of the first five responses, ChatGPT will tailor the output to your needs. If you're using it for strategy or planning, it's going to gather crucial context. Every answer is better with more information from you.

Get intentional with your prompting: add 4 simple words

Four simple words will improve your output from ChatGPT or Claude. Get even more specific by asking for ten questions. Then talk through your answers instead of typing them. Verbalising your thoughts lets you go off on tangents that give useful snippets of information for ChatGPT to take on board. Talking through your answers will give you as much clarity as it gives ChatGPT.

Use Wispr Flow to talk through your reasoning and answer its questions. Talking 200 words is far easier than typing. Faster too. The output is likely to be 10x what it would have been without that step. Submit your answers, let the robot think, then see how much richer the answer becomes when it's got all that extra information.

You have the answers. ChatGPT doesn't. Don't let it fill in the gaps. Get intentional with your prompting. Tell it to ask you questions first.

Access my best ChatGPT prompts .