Large language models (LLMs) are driving significant advancements across various sectors. Combined with robotics, they can lay the foundation for a new paradigm in healthcare, particularly in elderly care and companionship. For years, research has focused on the concept of carebots and robot companions. The latest capabilities of LLMs and robotics promise to address many challenges regarding interaction and personalized experience faced by earlier generations of companion robots. However, the potential of these emerging technologies is accompanied by significant ethical, technical, and social challenges. This chapter presents the design of an elderly care robot that exploits the capabilities of the latest generation of LLMs and considers the ethical, technical, and social implications for improving human-robot interaction capabilities. We report the technical design of a novel robotic platform for elderly care, detailing the devised interaction scenarios, a prototype implementation with NAO and PEPPER robots, and the results of a technical validation that shows the efficiency and feasibility of our system. This study serves as a guide to identify future needs, explore the impact of companion robots on older adults, and understand acceptance of these new approaches.