What you'll get
You will know how to steer a disappointing answer into a good one with short, specific feedback, instead of starting again.
The idea
The first answer is a starting point, not the final result. Most people either accept a mediocre first answer or give up and rephrase the whole question. The quickest route to something good is usually to tell the AI what is wrong, in plain words, and let it try again.
The explainer above shows this happening with a short LinkedIn post. The written version is below.
Feedback that works
Short, specific feedback works best. Say what to keep as well as what to change.
Good content, but half the length. Keep the first and last paragraphs as they are.
This sounds like a press release. Make it sound like me talking to a colleague: plainer words, shorter sentences, no exclamation marks.
It's too general. Use the example of our Leeds office move from my notes instead of the generic one.
Something about this doesn't work but I can't say what. Give me three quite different versions so I can compare.
When to start again
If you have given three or four rounds of feedback and it is getting worse, the conversation may have gone down the wrong path. Start a new chat, but bring what you learned: put the important points into your first message.
Take it to your AI
Copy this into Claude or ChatGPT to get started
This prompt sums up the guide and asks your AI to walk you through it using your own work. Paste it in, then answer its questions.
I'd like to practise giving you better feedback. Give me a first draft of a short LinkedIn post about [a topic from my work]. Don't try too hard on the first version. Then let me push back on it. After each of my replies, tell me what you changed and suggest one piece of feedback I could give next that would improve it even more. Stop when I say I'm happy.
Written for Claude and ChatGPT, Oct 2026
Was this guide useful?
Thanks, that helps us decide what to improve.