Commercial Authorship and Content Fraud Leave Distinct Bibliometric Signatures: a Retrospective Feasibility Study of the Author Research Integrity Risk Index (ARI²)

No validated author-level bibliometric tool currently exists to detect the traces that paper mills and authorship-for-sale networks leave behind as they compromise the scientific record. The Author Research Integrity Risk Index (ARI²) introduced here combines four public-record signals (network fragmentation, retraction burden, output anomaly, and citation integrity) into a composite score designed to flag the co-authorship and output anomalies that commercial-authorship practices tend to produce. The index was tested retrospectively in a feasibility study of 28 source-verified authors: 5 commercial-authorship (paper mill) cases, 10 content-fraud / research-record unreliability cases, and 13 comparison controls with verified Ret = 0. ARI² showed higher scores for paper-mill cases versus controls (median 76.4 vs. 31.8; U = 61, p = 0.003 on unrounded analytical scores; rank-biserial r = 0.877; AUC = 0.938, 95% CI 0.769–1.000). Content-fraud cases did not significantly differ from controls (U = 89, p = 0.145), and ARI² showed stronger within-cohort separation for commercial-authorship cases than for content-fraud cases. The retraction-free variant (ARI²-NR) did not reach conventional significance for any comparison (paper-mill cases versus controls: AUC = 0.800, p = 0.059), indicating that prospective pre-retraction discriminatory capacity remains unproven. In exploratory unadjusted analyses, the Per-Paper Geographic Dispersion Index (PPGDI) was higher among commercial-authorship authors than content-fraud authors (mean 2.56 vs. 1.10; U = 42, p = 0.007) and controls (mean 1.31; U = 60, p = 0.008), providing preliminary support for a prospective detection module. The paper-mill versus controls comparison was the only pre-specified primary endpoint; all remaining comparisons were exploratory and unadjusted for multiplicity. The underlying design question was whether different misconduct mechanisms leave different public-record traces. ARI² in its current form is retrospective by design: Retraction Burden overlaps with case ascertainment, and any future screening role will require prospective validation in field-specific cohorts, calibrated reference norms, and procedural safeguards against misuse before deployment. The index is openly implemented at kelleniindex.com.Associated data and reproducibility code are available at Zenodo DOI: https://doi.org/10.5281/zenodo.20044675. The all-versions concept DOI for the associated dataset/code record is: https://doi.org/10.5281/zenodo.20018816.