Automatic question generation (AQG) is a compelling and challenging area of research within Natural Language Processing (NLP). This field focuses on the automatic creation of questions from a given text, enhancing applications such as reading comprehension exercises, reducing the time teachers spend preparing questions and aiding second-language learners. The primary motivation for AQG research is the need for scalable, effective solutions to content production, evaluation and knowledge sharing. Traditional question-creation methods are labour-intensive and time-consuming, necessitating human annotators. With the rapid growth of digital data, automated systems capable of extracting relevant information and generating questions are increasingly essential. AQG aims to develop computational systems that can understand text and produce meaningful questions that test comprehension and problem-solving skills. This paper classifies various AQG approaches, analyses their results using automatic evaluation scores, reviews different datasets and their availability, and discusses current and potential evaluation techniques.