An assistant says a reply is ‘personalized for you.’ That label is less useful than knowing which part of the suggestion depends on your preferences or conversation history.
A practical explanation might say that the draft is brief because you selected a concise work style, or that it asks for a date because the conversation contains no confirmed date. Those explanations point to something you can inspect and change.
A vague statement such as ‘The system understands your personality’ does not give the same control. It can sound impressive while leaving the user unable to identify a mistaken assumption.
An explanation should also avoid presenting inferred feelings as facts. ‘The sender may be asking for reassurance’ is different from claiming to know their emotional state with certainty.
When reviewing personalized output, ask whether the explanation would help you repair a bad draft. Could you correct the relevant preference, remove an inappropriate source or supply the missing detail?
The most useful explanations are tied to the actual decision in front of the user. They do not need to expose every technical operation. They need to make the reason for the suggestion understandable enough that a person can decide whether it belongs in this conversation.