Add a field for evaluating whether the message sounds natural
As far as I can see, it is not possible to completely prevent the copy-pasting of the answers given by platforms such as ChatGPT (although it is stated that it is prohibited in the guide). These answers can sometimes be very obvious. On the other hand, it may be desirable for the assistant to speak in this way in various situations. For these reasons, I think that evaluating the naturalness of speech can be useful data.
Adding a field specifically for this situation can both make it easier to review messages that are likely to be copy-paste, and can be used to train more natural-speaking AI models.
or
In addition, if I remember correctly, the answers are planned to be created by artificial intelligence after a certain point within the scope of the project. The existence of such a field can also be used to test whether these messages are distinguishable by humans.