Contemporary societies struggle with Fake News, disseminated primarily through social media and newsrooms. The detection of fake news is therefore crucial, as it erodes trust in our democratic institutions, a key pillar of a functioning modern society. In this systematic literature review, we analyse AI solutions aiming at the highly accurate detection of fake news; our goal is to assess their techniques, as well as their ability to address key challenges, and to identify their strengths and limitations. Our findings reveal that Deep Learning dominates the field while other ‘classical’ Machine Learning solutions remain relevant. We also observe a shift towards novel AI solutions, such as Transformer-based models and hybrid architectures that combine elements from different AI paradigms. The research in the field is shifting towards metadata-agnostic solutions, focusing on the development of systems with advanced contextual analytical capabilities, taking significant care to counter potential problems of overfitting. Finally, a clear thematic focus of the data emerges, with the data pertaining to major social issues such as politics and the COVID-19 pandemic.