Artificial Intelligence and Machine Learning in the Transformation of Forensic Accounting: A Systematic Literature Review

This research analyzed the integration of artificial intelligence (AI) and machine learning in forensic accounting as tools for the detection of accounting fraud. Through a systematic review of the literature in Scopus under the PRISMA protocol, 76 documents published between 2019 and 2026 were selected and analyzed. The results show that supervised machine learning techniques, assembly methods, deep learning and natural language processing have demonstrated superior capacity to rule-based approaches to detect fraudulent patterns, when quality data and professionals capable of interpreting their results with forensic criteria are available. A recurring finding throughout the corpus is that machine learning models tend to perform better than rule-based approaches in contexts where sufficient quality data is available. Although this advantage is not universal, it depends on the type of fraud analyzed, the design of the model, the evaluation metrics used, and the institutional context of implementation. AI expands the practitioner’s ability to identify anomalous patterns in volumes of information that manual analysis cannot cover. However, significant barriers persist related to data quality, model interpretability, professional resistance to technological updating, and the absence of regulatory frameworks that grant evidentiary validity to algorithmic results. Latin America and Africa show a low investigative production in this field, despite concentrating high levels of vulnerability to financial fraud. The study concludes that the transformation of forensic accounting using AI is a social, institutional and educational process, not exclusively technical, which requires interdisciplinary training, data governance and specific ethical frameworks for the forensic field.