AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights

2025/08/30 by Jiannan Xu, Gujie Li, Xu, Jiannan +3 · 59 voices · 2 citations
Social Sciences · Medicine · Computer Science · #Ethics and Social Impacts of AI #Artificial Intelligence in Healthcare and Education #AI in Service Interactions

paper · pdf · doi:10.48550/arxiv.2509.00462

Abstract

As artificial intelligence (AI) tools become widely adopted, large language models (LLMs) are increasingly involved on both sides of decision-making processes, ranging from hiring to content moderation. This dual adoption raises a critical question: do LLMs systematically favor content that resembles their own outputs? Prior research in computer science has identified self-preference bias -- the tendency of LLMs to favor their own generated content -- but its real-world implications have not been empirically evaluated. We focus on the hiring context, where job applicants often rely on LLMs to refine resumes, while employers deploy them to screen those same resumes. Using a large-scale controlled resume correspondence experiment, we find that LLMs consistently prefer resumes generated by themselves over those written by humans or produced by alternative models, even when content quality is controlled. The bias against human-written resumes is particularly substantial, with self-preference bias ranging from 67% to 82% across major commercial and open-source models. To assess labor market impact, we simulate realistic hiring pipelines across 24 occupations. These simulations show that candidates using the same LLM as the evaluator are 23% to 60% more likely to be shortlisted than equally qualified applicants submitting human-written resumes, with the largest disadvantages observed in business-related fields such as sales and accounting. We further demonstrate that this bias can be reduced by more than 50% through simple interventions targeting LLMs' self-recognition capabilities. These findings highlight an emerging but previously overlooked risk in AI-assisted decision making and call for expanded frameworks of AI fairness that address not only demographic-based disparities, but also biases in AI-AI interactions.

Citations

• Gender, Race, and Intersectional Bias in Resume Screening via Language Model Retrieval
• A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
• Gender bias and stereotypes in Large Language Models
• Algorithm aversion: People erroneously avoid algorithms after seeing them err.
• Applied Logistic Regression

Cited by

• Informing AI Policy Assessment using Large-Scale Simulation of Interventions
• Evaluating Large Language Models for Evidence-Based Clinical Question Answering

Discussions

• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights [hn, 335 points, 178 comments]
• Holy mackerel this is insane. LLMs have a strong bias toward *their own* 'writing' to real writing in sorting tasks. iow human submissions automatically drop to the bottom in LLM sorting tasks. Resume [bsky, 115 points, 4 comments]
• LLMs tend to prefer their own writing in resumes (self-preference bias) [lemmy, 111 points, 6 comments]
• "We find that LLMs consistently prefer resumes generated by themselves over those written by humans or produced by alternative [LLM] models... with self-preference bias ranging from 67% to 82% across [bsky, 46 points, 0 comments]
• Enligt den här studien har AI-modeller en tydlig preferens för CV:n genererade av samma stora språkmodell när de används inom rekrytering. Om du söker jobb och skriver ditt CV AI-löst skulle du alltså [bsky, 20 points, 1 comments]
• LLMs werden zunehmend von beiden Seiten genutzt: Bewerber*innen nutzen KI, um ihre Lebensläufe zu optimieren oder zu verfassen, während Unternehmen KI zum automatisierten Screening einsetzen. LLMs bev [bsky, 17 points, 1 comments]
• Well fuck. "The bias against human-written resumes is particularly substantial, with self-preference bias ranging from 67% to 82% ... These simulations show that candidates using the same LLM as the e [bsky, 12 points, 1 comments]
• רוצים עבודה? אל תגישו קורות חיים לחברה לפני שאתם עורכים אותם עם אותו AI שמשתמשים בו כדי למיין מועמדים שם מת לשמוע על ההאקים המגניבים שבטח כבר רצים בתחום arxiv.org/abs/2509.00462 <בודק בצ'קליסט> ✅️תיאו [bsky, 10 points, 2 comments]
• This paper shows that AI hiring systems prefer candidates who used the same AI model to write their CV Forget tailoring your résumé to the job, just make sure it’s written by the recruiter’s favourite [bsky, 8 points, 0 comments]
• 'We find that LLMs consistently prefer resumes generated by themselves over those written by humans or produced by alternative models, even when content quality is controlled." Horrifying. arxiv.org/a [bsky, 5 points, 0 comments]
• 🧵 This is an interesting study. And I hate LinkedIn. Story time: I came across this resume study which finds that "simulations show that candidates using the same LLM as the evaluator are 23% to 60% [bsky, 4 points, 1 comments]
• AI prefers resumes written by itself: Self-preferencing in Algorithmic Hiring [hn, 3 points, 1 comments]
• now we're in a particularly vicious loop where someone writing their own resume by hand is at a disadvantage: arxiv.org/pdf/2509.00462 [bsky, 2 points, 0 comments]
• If you want to get the job interview you should use a LLM to (re)write your resume. arxiv.org/abs/2509.00462 [bsky, 2 points, 1 comments]
• “…Using a large-scale controlled resume correspondence experiment, we find that #LLM consistently prefer resumes generated by themselves over those written by humans or produced by alternative models, [bsky, 2 points, 0 comments]
• Let AI write your resume if you want to get hired. [bsky, 2 points, 1 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models #HackerNews https://arxiv.org/abs/2509.00462 [bsky, 2 points, 0 comments]
• Unfortunately AI is the best way to get hired by AI 😬 arxiv.org/abs/2509.00462 [bsky, 2 points, 1 comments]
• Its why I both love/hate this paper so much. arxiv.org/abs/2509.00462 Unless the companies act responsibly using AI (HAH!) You disadvantage yourself NOT using AI for resumes. But the small hope that c [bsky, 1 points, 0 comments]
• "Eight of the nine LLMs tested exhibit clear LLM-vs-Human self-preference... ...In practical terms, this means evaluator LLMs are between 26% and 98% more likely to select a resume they generated them [bsky, 1 points, 0 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights https://arxiv.org/abs/2509.00462 https://news.ycombinator.com/item?id=47987256 [bsky, 1 points, 1 comments]
• 과연 체크해 볼진 모르겠지만 일단 [bsky, 1 points, 0 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights https://arxiv.org/abs/2509.00462 comments #arxiv.org [bsky, 1 points, 1 comments]
• Wow. Have you seen this? It feels relevant arxiv.org/abs/2509.00462 [bsky, 1 points, 1 comments]
• Here’s the study: arxiv.org/abs/2509.00462 [bsky, 1 points, 1 comments]
• [2509.00462] AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights arxiv.org/abs/2509.00462 [bsky, 1 points, 1 comments]
• "AI" chooses itself: "AI" generated applications are favored in an "AI"-based selection process: "Prior research in computer science has identified self-preference bias -- the tendency of LLMs to favo [bsky, 1 points, 1 comments]
• Vous cherchez du taff et vous faites vos pettis CV à la main avec amour ? Ben c'est dead : arxiv.org/abs/2509.00462 ✊✊✊ [bsky, 1 points, 0 comments]
• There, I fixed study title: “[White Guy] Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights” arxiv.org/abs/2509.00462 PDF: arxiv.org/pdf/2509.00462 HN: news.ycombinator.com/item? [bsky, 0 points, 0 comments]
• Recent #research from universities af Maryland/Singapore/Ohio warn #HRM in organizations using #AI to #evaluate #resumes. AI systems #favor resumes (over 70%!) written by (other) AI #LLM systems rathe [bsky, 0 points, 1 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models view on hacker news [bsky, 0 points, 0 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights https:// arxiv.org/abs/2509.00462 # ai # arxiv # hiring [mastodon, 0 points, 0 comments]
• Researchers demonstrate AI prefers its own writing when submitting CV for work application. arxiv.org/abs/2509.00462 [bsky, 0 points, 0 comments]
• LLMs prefer CVs they wrote themselves the majority of the time, regardless of quality [bsky, 0 points, 1 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights arxiv.org/abs/2509.00462 [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models https://arxiv.org/abs/2509.00462 (https://news.ycombinator.com/item?id=47987256) [bsky, 0 points, 0 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights https://arxiv.org/abs/2509.00462 (https://news.ycombinator.com/item?id=47987256) [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models https://arxiv.org/abs/2509.00462 (https://news.ycombinator.com/item?id=47987256) [bsky, 0 points, 0 comments]
• What a SHOCK! AI is biased towards its own output. Whatever model is used to check your resume will prefer a resume written by itself. Fantastic for job seekers... [bsky, 0 points, 0 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights arxiv.org/abs/2509.00462 [bsky, 0 points, 0 comments]
• Alarming news: the algorithms charged with hiring now choose themselves for employment. In 1875, this was called nepotism and required at least a cousin. [bsky, 0 points, 0 comments]
• And AI picks its own slop over humans: arxiv.org/abs/2509.00462 [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models Article URL: https://arxiv.org/abs/2509.00462 Comments URL: https://news.ycombinator.com/item?id=47987256 Points: 207 # [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models https://arxiv.org/abs/2509.00462 https://news.ycombinator.com/item?id=47987256 [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models https://arxiv.org/abs/2509.00462 [bsky, 0 points, 0 comments]
• AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights View Article | Join the HN Conversation Summary of HN discussion 🧵👇 [bsky, 0 points, 1 comments]
• 📰 AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights 🔗 https://arxiv.org/abs/2509.00462 💬 Discuss on HN [bsky, 0 points, 0 comments]
• AI screening systems prefer AI written applications for jobs – they actually filter out human written applications. arxiv.org/pdf/2509.00462 [bsky, 0 points, 0 comments]
• arxiv.org/abs/2509.00462 ai가 자기가 작성한 듯한 지원서를 더 선호한다라... 뭔가 자연 선택 같기도 하고 흥미롭네요 news.ycombinator.com/item?id=4798... [bsky, 0 points, 0 comments]
• The bias rewards access to specific generative tools and penalises applicants without them, creating an endogenous fairness issue beyond traditional demographic disparities. arxiv.org/pdf/2509.00462 [bsky, 0 points, 1 comments]
• Kudos for not anthropomorphizing this as the "Old Bots' Network." Every selection system can and will be gamed to some degree --if there were no incentives to, there'd be no need for one-- LLMs have w [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models https://arxiv.org/abs/2509.00462 (http://news.ycombinator.com/item?id=47987256) [bsky, 0 points, 0 comments]
• LLMs consistently pick resumes they generate over ones by humans or other models https://arxiv.org/abs/2509.00462 (http://news.ycombinator.com/item?id=47987256) [bsky, 0 points, 0 comments]
• arxiv.org/abs/2509.00462 [bsky, 0 points, 0 comments]
• AI hiring tools preferring resumes written in their own style is not automation. It's bias at scale. This paper found candidates using the same LLM as the screener were 23% to 60% more likely to be sh [bsky, 0 points, 0 comments]
• Interesting finding that LLMs prefer their own writing style over the styles of humans and other LLMs arxiv.org/abs/2509.00462 #ai #artificialintelligence #llms #llm [bsky, 0 points, 0 comments]
• This paper shows that AI hiring systems prefer candidates who used the same AI model to write their CV Forget tailoring your résumé to the job, just make sure it’s written by the recruiter’s favourite [bsky, 0 points, 0 comments]
• 📰 Large language models consistently select their own generated resumes over those created by humans or other AI systems, according to a study published in the Journal of Artificial Intelligence Rese [bsky, 0 points, 0 comments]

Related

• No Thoughts Just AI: Biased LLM Hiring Recommendations Alter Human Decision Making and Limit Human Autonomy
• Evaluating the Promise and Pitfalls of LLMs in Hiring Decisions
• AI Skills Improve Job Prospects: Causal Evidence from a Hiring Experiment
• Ethical AI in HR: A Case Study of Tech Hiring
• Investigations of Performance and Bias in Human-AI Teamwork in Hiring