Artificial intelligence (AI) in higher education has proven to be a useful learning technology; it can help learners achieve positive learning outcomes in the learning environment and can also enable teachers to better understand learners' learning status and further improve their teaching strategies. This study reviewed the top 50 AI in higher education studies in the Web of Science database from the perspective of highly cited papers and based on a technology-based learning model. The results show that predictions of learners' learning status (including dropout and retention, student models, and academic achievement) are most frequently discussed in the AI in higher education studies; AI technology is most commonly applied in engineering (including computer courses); AI technologies most often play the role of profiling and prediction in higher education, followed by intelligent tutoring systems and assessment and evaluation. In terms of research issues, the most frequently discussed issues are learning behaviour, accuracy, sensitivity and precision, cognition and affect. Learners' higher order thinking skills, collaboration or communication, self-efficacy or confidence and skills are less frequently discussed in AI in higher education studies. Accordingly, we propose potential research issues and practitioner notes for AI in higher education as a reference for researchers, educators and decision-makers.