Taxonomy-Based Coverage Benchmarking: A Structural Audit Framework for AI-Driven Classification Systems

This dataset contains the complete criterion-level structural audit mapping of six AI-driven journal classification systems (OpenEvidence, Perplexity, Claude, ChatGPT, Gemini, Grok) against a 59-item structured reference taxonomy derived from the 3rd edition of Beall’s criteria. Each criterion was coded as explicitly included (1.0), partially addressed (0.5), or not addressed (0.0). Both categorical (Yes/Partial/No) and numeric encodings are provided. The dataset includes: Criterion-level mappings (Mapping_Data sheet) Domain-level and global coverage calculations (Domain_Coverage sheet) Documentation of coding structure and formulas (README sheet) Standardized query protocol used for data extraction Global coverage values range from 40.7% to 98.3%, reflecting variability in structural operational alignment across systems. Data collection date: 13 February 2026. This dataset represents a time-stamped structural audit snapshot. AI systems are dynamic and may generate different outputs over time.