Teaching tax-related regulations have always been a challenge due to the inclusion of external variables that hinder the learning process, such as the high complexity of tax systems and legislation variability. Universities have sought different alternatives to support the teaching process both outside and inside the classroom, ranging from recreational activities to active learning. This article will show the experience resulting from using a chatbot to support learning in accounting students for the teaching of tax regulations related to the Chilean tax system, comparing two types of tools, on the one hand, a free conversation chatbot using natural language processing versus a rule-based chatbot driven by a decision tree. The experimentation process was carried out with 50 higher education students, divided into an experimental group and a control group, in two different courses. The results obtained demonstrated in both cases greater effectiveness of the use of the chatbot in learning the tax matter, both in the free conversation chatbot where the experimental group obtained a 15.7% improvement versus the control group that obtained a 1.05% improvement, as in the chatbot that applied decision tree where the experimental group obtained a 32% improvement versus the control group with 5.2%. Considering the complexity of the content in tax matters and the little innovation in the existing teaching subjects in the area and that the students improve learning using both chatbot tools compared to other learning techniques, is considered a relevant contribution to teaching innovation.