Emotional AI represents a relatively new research area, and to date, its focus has been on the development of AI methods to recognise instantaneous human states, for example, whether a person is currently happy or experiencing acute stress. However, because the majority of emotion detection studies to date have been conducted in laboratory settings, many technologies are not yet sufficiently mature for real-world applications. For example, though truly emotionally intelligent AI should be convenient for the users, adaptive to their personalities (given that the expression of the same emotion can vary across individuals), and capable of also assessing long-lasting human conditions, such as chronic stress, research into these problems is only starting. This chapter offers an overview of the most popular data sources and AI methods and discusses the main challenges that must be addressed to develop emotional AI methods for long-term real-world use.