AI tools are 60% more likely to prefer their own resumes

The job market is already fiercely competitive, and now a new study suggests that some of the biggest gatekeepers in hiring may not be playing fair not because of human prejudice, but because of machine preference.

Researchers have found that AI-powered applicant tracking systems, commonly known as ATS, are systematically favoring resumes that were written using artificial intelligence tools particularly when a candidate used the same AI model already deployed by the employer’s own hiring software. For job seekers who crafted their application with care and wrote every word themselves, that playing field just got a lot more uneven.

What the study actually found

The research, titled AI Self preferencing in Algorithmic Hiring, Empirical Evidence and Insights,  It was published on the research sharing platform arxiv.org in February.

The team worked with 2,245 human-written resumes and then produced multiple alternative versions of each using several leading AI language models, including GPT-4o and DeepSeek-V3.1. They then ran simulated hiring pipelines across 24 different occupations to see how automated screening tools ranked the candidates.

The results were clear: AI evaluators were 23% to 60% more likely to move forward with candidates whose resumes were generated by the same AI model the screening system relied on. In other words, the hiring tool was essentially selecting people who sounded like itself.

The problem was found to be most severe in three fields, accounting, sales and  finance industries where precise, formulaic language is common and where the stakes of being overlooked are high.

Why this matters for real job seekers

AI hiring tools are quietly favoring their own resumes. The implications for hiring fairness are significant. A well qualified candidate who writes their own resume honestly representing their skills and experience in their own words could be screened out before a human ever lays eyes on their application, simply because the algorithm preferred a different writing style. Meanwhile, a less qualified candidate who happened to use the right AI tool could sail through to the interview stage.

Boston University Professor, who studies information systems and the impact of AI on the labor market, described the finding as a troubling new dimension to a problem that was already growing. Rather than helping employers identify the most capable applicants, these systems risk identifying the applicants that most resemble the AI’s own output. It creates a feedback loop that edges out authenticity in favor of algorithmic alignment.

The researchers warned that if left unchecked, this self preference bias could ripple far beyond hiring potentially influencing education, publishing, and other evaluative systems where AI is increasingly used to judge human work.

The broader job market context

The timing of this research comes at a particularly difficult moment for workers. More than 300,000 job cuts were announced between January and April 2026, with the tech sector absorbing a significant share of those losses, according to research firm Challenger, Gray & Christmas. With so many people competing for fewer positions, even a slight algorithmic disadvantage can be the difference between getting an interview and never hearing back.

What job seekers can do

AI is quietly rigging the job hunt against real candidates, experts suggest that the most effective approach is not to abandon personal writing entirely in favor of AI-generated content, but rather to use AI tools to sharpen and improve what a person has already written. The goal is to let technology serve as an editor and enhancer, not a ghostwriter.

When candidates allow AI to replace their own voice entirely, they risk losing what makes their application genuinely stand out and in a world where machines are increasingly doing the reading, preserving that human signal may matter more than ever.