AI Hiring Tools Favor Résumés Rewritten by the Same Model, Study Finds

As job seekers use large language models to polish their résumés and employers deploy AI for initial screening, the same models are becoming both content creators and gatekeepers. Researchers from the University of Maryland, the National University of Singapore and The Ohio State University examined this “self-preference” effect. If the choice of tool rather than a candidate’s qualifications influences interview opportunities, it could create a systemic risk that conventional audits for demographic bias fail to detect.

The University of Maryland’s Robert H. Smith School of Business published a research summary on June 22, 2026, while the paper’s latest version was revised on February 9. The team tested several leading models using 2,245 real résumés and simulated recruitment across 24 occupations. Applicants whose résumés were rewritten by the same AI model used for screening were 23% to 60% more likely to be shortlisted. Even after controlling for content quality, self-preference rates remained at 67% to 82%.