Supplementary data for "Artificial Intelligence Methods for Automating Literature Review Workflows: A Critical Review"

Supplementary data for the critical review "Artificial Intelligence Methods for Automating Literature Review Workflows," which analyses 191 publications from 2016 to 2025 and maps methods, evaluation practices, and validation maturity across seven literature-review workflow phases. This deposit contains the complete study-level analytical record underlying every count reported in the article. It comprises a Supplementary Data workbook with 191 records coded across 42 fields (bibliographic metadata, publication and study-design classification, workflow phases, exact methods and named systems, technique families, preprocessing, training and test data, evaluation metrics, reported results, resource availability, validation tier, and stated challenges); a condensed included-publication ledger (Supplementary Table S1, 191 records); a full-text exclusion ledger with standardized reason groups (Supplementary Table S2, 60 records); a list of reports sought but not retrieved (Supplementary Table S3, 29 records); and the Supplementary Methods and Coding Framework defining eligibility criteria, controlled variables, validation-maturity tiers, metric categories, and counting rules. Three conventions govern how these files reconcile with the totals in the article. The primary phase is mutually exclusive, so phase totals use a denominator of 191. Technique families and metric categories are nonexclusive and include only methods that were applied or evaluated, giving a denominator of 170; percentages in those tables do not sum to 100 percent. Retrieval-augmented generation is recorded as a distinct architecture, and a RAG system using a generative model contributes to both the RAG and the LLM/generative-AI categories. Validation tiers describe the strength of the evaluation design only and are not risk-of-bias or overall quality scores. Records were assembled from two searches, an initial search covering 2016 to March 2024 and an update covering March 2024 to December 2025. One reviewer conducted screening, eligibility assessment, and study-level coding, supported by structured decision rules and automated consistency checks. The files therefore reflect single-reviewer coding rather than independent duplicate coding. This deposit contains no new primary data; all records derive from previously published literature.