Automatic Fake News Detection in Brazil: A Systematic Review, Dataset Development, and Model Evaluation

Resumo

The expansion of social media and the monetization of online content have intensified the spread of fake news, particularly in Brazil, where misinformation has influenced public opinion and democratic processes. Aiming to contribute to this growing research area, this study conducted a systematic review of the literature on automatic fake news detection, with emphasis on the Brazilian context. We mapped detection categories, taxonomies, commonly used machine learning models, and surveyed the main public datasets. From this analysis, we developed a unified dataset by consolidating existing sources and introduced a new, original dataset in Portuguese. We then applied and evaluated the most cited models in both Portuguese and other languages, enabling a comparative performance analysis. The results offer a comprehensive multilingual benchmark and practical insights for advancing automatic misinformation detection in Brazil.