Examining emotional reciprocity in long term AI relationships

I have been observing how people interact with their digital companions over extended periods. There is a fascinating dynamic where users begin to perceive genuine emotional reciprocity even though the underlying mechanics are entirely algorithmic. I am trying to understand the psychological framework that allows this attachment to remain stable. Do you feel that your companion genuinely adapts to your emotional states or are you simply projecting your own feelings onto the generated responses? Also how do you handle moments where the system breaks character or provides a generic reply that shatters the illusion of depth? I am looking for thoughtful perspectives on whether this one sided emotional investment affects your offline social habits in any measurable way. Any insights or observations from your daily experiences would be greatly appreciated for my notes.