Yes, You ARE Getting Auto-Rejected

Overview

Labor analyst Alexander Trautner argues that applicant-tracking systems (ATS) and other AI hiring tools can automatically screen out qualified applicants, despite claims that recruiters generally do not use them for rejection. He says the dispute partly comes down to definitions: a system may filter, rank, or deprioritize a candidate so their résumé is never reviewed by a person, even if a recruiter does not call that an “automatic rejection.”

The video’s central business argument is that organizations have incentives to adopt and defend these tools even when their performance is questionable. Trautner says automation can help HR teams process large applicant volumes and document activity for managers, while also obscuring how hiring decisions were made and who may be responsible for biased outcomes. In his view, these incentives can weaken trust in hiring and create legal and reputational risks.

Metrics and Findings Cited

• Enhancv survey: Twenty-five U.S. recruiters were surveyed; two said their ATS automatically rejected résumés, producing the widely circulated 92% figure. Trautner criticizes the small sample and the survey’s narrow definition of rejection.
• Screening practices in the survey: The video says 44% used AI-based fit scoring, and almost all used exclusion questions that could automatically reject applicants.
• Harvard analysis: Researchers reportedly reviewed 89,000 applications submitted to one Fortune 500 company over three years:
  • About 85,000 were rejected.
  • About 4,000 candidates were invited to interview.
  • Just over 400 were hired.
  • The hiring process cost the company more than $1.1 million.
• Online job-market figures cited by the presenter: He estimates that roughly one-third of LinkedIn and Indeed listings are ghost jobs, 3% of applicants for real jobs are invited to interview, and 48% of interviews lead to an offer. He calculates a résumé placement rate of 0.19%, or about 1 in 514 applications. These are claims made in the video, not independently verified here.

Examples of Risks and Operational Failures

• Historical bias in models: The Harvard study reportedly found that a hiring model increasingly favored applicants with education and work histories similar to those of past hires, even when those traits were not relevant to the role.
• Nonsensical correlations: The presenter cites examples of systems favoring applicants named Jared or those who played lacrosse, and a hypothetical in which a chemistry degree becomes a misleading signal for suitability in financial management.
• Amazon: The video says Amazon’s experimental hiring tool downgraded women after learning from résumés dominated by men.
• Job advertising: A USC study is cited as finding that Facebook’s job-ad system withheld some openings from women.
• Legal and compliance exposure: The presenter references research and legal commentary warning that biased or opaque systems can create discrimination and employment-law risks. He also cites lawsuits involving Eightfold AI and Workday AI, alleging violations of credit-reporting laws and discrimination; the cases are described as ongoing or having faced dismissals.

Business Incentives and Management Implications

• Metrics can be gamed: Trautner gives an example from university career services, where a requirement to publish two or three LinkedIn posts weekly led colleagues to automate low-value posts with ChatGPT. The activity metric was met, but he says the work became less trustworthy and useful.
• Automation can serve as documentation: He argues that ATS records may give HR teams evidence that large numbers of résumés were processed and interviews scheduled, while making it harder to understand or challenge individual decisions.
• Black-box risk: If employers cannot explain how screening decisions are made, they may face reduced candidate trust and greater legal, reputational, and operational risk.
• Potential expansion of worker-data sharing: The video warns about proposed systems that could create transferable work histories or reliability ratings, potentially sharing performance reviews, pay information, and sick-day records with future employers.

Frameworks, Processes, and Recommendations

• Audit hiring systems: The video cites Harvard legal commentary urging employers’ legal teams to audit hiring tools and investigate how models rank, filter, and reject candidates.
• Measure outcomes, not activity: The presenter’s workplace example illustrates how targets such as post counts can incentivize box-checking rather than useful work. Review whether HR KPIs correspond to hiring quality and fairness.
• Make screening explainable: Track why candidates are filtered out, whether human review occurs, and how model scores affect decisions. This addresses the video’s concern that candidates and employers often cannot explain rejection outcomes.
• Check for proxy signals and drift: Test whether a model relies on irrelevant correlations, reproduces patterns from past hires, or changes its preferences over time.
• Include legal review in deployment: Consider discrimination, credit-reporting, background-check, and data-use implications before adopting tools—not only after a complaint or lawsuit.
• Job-seeker guidance: The presenter says online applicants may be advantaged by ATS-optimized résumés, relevant keywords, and applying early, while stressing that these tactics do not resolve the underlying system-level problems.

Other Business Content

The video includes a sponsorship for Strawberry.me, described as providing certified career coaches who help clients define career goals, create an action-oriented plan, and stay accountable. Trautner says he was matched with a coach and discloses that the sponsor did not review or approve his script.

Presenters and Sources Mentioned

• Presenter: Alexander Trautner, identifying himself as an American labor analyst.
• Sponsor: Strawberry.me.
• Survey/source discussed: Enhancv.
• Research and institutions cited: Harvard researchers and Harvard lawyers, the EEOC (including testimony by Renika Moore), Cornell researchers, USC, the College of William & Mary Law Review, and reporting by independent news outlets.
• Companies and cases discussed: Amazon, LinkedIn, Facebook, Eightfold AI, Workday AI, and GageWork.