This chapter explores techniques for crafting effective AI interactions through the assignment of personas. Unlike humans, who intrinsically develop common frames of reference, AI’s lack innate context, with language models like ChatGPT acting in ways that could be thought of as resembling “prediction machines”. As such, they are somewhat limited to providing users with what they calculate to be the statistically most reasonable response. Consequently, outputs may miss user intent if insufficient framing is supplied. By detailing parameters such as background and expertise when prescribing AI roles, users can create filters that steer responses towards heightened relevance. This chapter explains persona dimensions through examples of mixing and matching specialty skills, disciplines, experience levels, voices, and eccentricities to create unique profiles. It also discusses the process of designing an ideal AI research assistant and provides practical activities for readers to apply these concepts. Ethical considerations are addressed, with cautions against the non-consensual use of real identities and recommendations for using fictional characterisation. This chapter concludes with reflection questions to deepen understanding and stimulate further discussion on the effective and ethical use of AI personas in academic research.