Pet separation anxiety is an inevitable challenge in the bond between pets and their owners. Our study begins by analyzing existing smart pet devices that use AI and IoT for remote monitoring, behavior analysis, and emotional engagement. However, these devices often exhibit design biases, offering only mechanistic, pre-programmed responses without nuanced emotional support tailored to individual pets’ needs. This foundation frames our inquiry, as we apply design research methods to better understand how experts working in the pet care industry alleviate separation anxiety. Drawing on these insights, we present an AI-driven wearable prototype that utilizes machine learning to analyze physiological data and generate customized feedback. Through deep learning, the system optimizes its intervention mechanisms, offering support aligned with pets’ emotional and health needs. In addition to the technical design, our research underscores the ethical challenges in deploying AI for emotionally sensitive pet interactions, with a focus on algorithmic transparency and fairness in behavior classification. We emphasize the importance of ensuring device feedback does not misinterpret behaviors, preventing increased anxiety due to misclassification. This study provides a framework for next-generation smart pet devices, focusing on transparency, fair classification, and ethical standards to strengthen pet-owner bonds and support pet well-being.