Payroll‑fraud screening requires uncovering hidden employee‑payee links while distinguishing suspicious from legitimate relationships. We formulate the task as inductive link prediction and entity ranking on synthetic human‑resources and enterprise‑resource‑planning (HR/ERP) graphs, each containing unseen organizations and a matched benign twin for every planted case.\ \ On 24 held‑out organizations and a prospectively frozen extension, we compare analyst rules, explicit path features, and graph neural networks (GNNs). A gradient‑boosted model using chained‑join features achieved the highest overall link average precision (AP) of 0.823, surpassing the GraIL‑style GNN’s 0.663; its 4‑hop AP was 0.646 versus the GNN’s 0.272, indicating that multi‑hop gains are not specific to GNNs.\ \ In a new synthetic cohort with 2.0% positive entities, chained‑join boosting again led (AP 0.798), and the relational‑entity GNN also outperformed record‑aware rules (0.707 vs 0.504). With a 2% review budget, their precisions were 0.695 and 0.602 respectively.\ \ When 40% identifier joins failed, the entity GNN ranked above chained‑join boosting (AP 0.712 vs 0.683, descriptively). Structure‑only auditing exposed a shortcut in the initial generator; matched twins reduced the audited twin‑pool AUROC to 0.493‑0.514.\ \ These results separate multi‑hop candidate discovery, record‑based discrimination, and review prioritization within an explicit synthetic evaluation design. Code and data are available from the authors on request.\ \ Review: The study provides a clear comparative view of rule‑based, path‑feature, and GNN approaches for payroll fraud detection on synthetic HR/ERP graphs, highlighting the importance of multi‑hop information and showing that GNNs are not the sole beneficiaries of such structure.