One of the uses of my many AI assistants is to help me improve my own posts. For example, I ask if I have thought about the most important topics and questions, if the whole thing is logically structured and written in an understandable way.
AI tools are actually quite good at this. They have broad knowledge and are specialists in analyzing text and other data.
When it comes to feedback, the only thing standing in their way is: They are too well trained. Literally.
As part of their training process, these AI applications learn which responses are desirable and which are not. They are often trained to be extra polite to the user. They are also meant be positive, motivating, and supportive.
But anyone who knows what makes feedback helpful understands: It’s not enough to just point out what’s good. You also need to be clear about where and how something can be improved.
But AI assistants are sometimes reluctant to do this or give you an incomplete answer.
To get around this, you can ask your assistant explicitly to give you honest, unvarnished feedback.
Maybe even make it a game: the more suggestions it has, the better. After all, the AI wants to fulfill your wishes as well as possible. You can take advantage of that.
Another option I recently read about on Reddit: Pretend that your text was actually written by someone else. This little twist can already release the honesty brake that the training has installed. After all, the AI no longer has to treat you nice and friendly, because it’s talking about someone else’s text, right?
And as a bonus tip: If you want general feedback, make sure your prompt is as open-ended as possible and doesn’t give any direction. Remember: Your prompt determines which answers are possible.
For example, do not tell the AI in this case how you are not satisfied with the introduction or that you suspect the language level is not appropriate. Since the AI is calibrated to follow you, it will most likely confirm whatever you tell it.
Just asking for general suggestions for improvement instead, leaves a lot of options open.