Artificial Intelligence, Machine Learning and Digital Forensics in Forensic Quality Assurance: A Systematic Review

Background. Artificial intelligence is being adopted across forensic practice faster than the quality assurance frameworks governing forensic reporting have adapted to it. Whether reported algorithmic performance has been established within those frameworks has not previously been assessed systematically. Objectives. This systematic review synthesises evidence on artificial intelligence, machine learning and digital forensic technologies in relation to forensic quality assurance across all forensic subdisciplines and jurisdictions. It addresses four questions: what performance AI-enabled forensic systems report and under what evaluation conditions; the extent to which that performance has been established within accreditation and method-validation frameworks; how the evidence is distributed across forensic domains, technologies and stages of operational translation; and what governance, standardisation and equity barriers the evidence identifies. Methods. Conducted in accordance with PRISMA 2020 and reported following SWiM. Two search waves covered publications from 2014 to July 2026, combining four concept blocks (forensic science; quality assurance and accreditation; artificial intelligence and machine learning; digital and cyber forensics) across Scopus, Web of Science, PubMed, IEEE Xplore, ScienceDirect and Google Scholar, supplemented by backward citation tracking. Sixty studies were included and appraised using the Mixed Methods Appraisal Tool 2018. Every included record was verified line by line against its source document. Substantial heterogeneity in design, technology, domain and outcome definition precluded meta-analysis; findings were synthesised narratively using pre-specified, reproducible counting rules for evidence gap mapping, technology mapping, operational readiness classification and directional trend analysis. Registration status. Registered retrospectively. PROSPERO registration was not available as the review's outcomes are not health-related and therefore fall outside that register's scope. The full protocol is attached.