AI Hiring Bias Tracker

AI resume screeners shown 85% of shortlisted candidates were white, 9% were Black, in controlled lab study

Résumés with Black-associated names required 50% more applications to get a callback than identical résumés with white-associated names

LinkedIn algorithmic job-matching showed Black and Hispanic professionals were systematically recommended for lower-wage roles than white peers with identical qualifications

EEOC filed first AI-bias hiring charge against iTutorGroup: platform automatically rejected applicants based on age (40+)

Amazon scrapped internal AI recruiting tool after it downgraded résumés containing the word "women" (e.g., "women's chess club")

UW AIES study: AI model shown only male candidates for 52% of shortlisted slots; female candidates received 11%

Mobley v Workday (class cert 2024): 200M+ workers over age 40 potentially affected by AI screening that used protected characteristics as proxies

HireVue facial analysis feature was quietly dropped in 2021 after bias audits found it encoded race and gender signals from facial geometry

EEOC 2022 technical assistance: AI tools that use protected-characteristic proxies (zip code, name, graduation year) constitute illegal discrimination under Title VII and ADEA

Pymetrics (now Harver) commissioned audit by Bain found race-based performance gaps of 4-6 percentage points in neurocognitive game scores

Applicants with disabilities are 26% less likely to receive a callback when AI screening detects assistive-technology signals in résumé metadata

AI video interview platforms scored non-native English speakers 12% lower on "communication" dimensions than native speakers with identical answer content

Only 18 US states had any AI hiring-transparency law as of Q1 2026; no federal statute specifically regulates algorithmic hiring decisions

New York City Local Law 144 (2023) requires annual third-party bias audits for automated employment decision tools used in NYC

Illinois Artificial Intelligence Video Interview Act (2020) requires employer disclosure and candidate consent before using AI to analyze video interviews

38% of job seekers globally reported using AI tools to optimize their résumés or applications as of 2023

Harvard Kennedy School: algorithmic hiring tools screen out approximately 27% of job-ready workers flagged as "non-traditional" by automated filters

Workday face-value audit (2022) found adverse impact ratios below 4/5ths rule for Black applicants in screening scores at 3 of 10 major enterprise clients

EU AI Act (2024): automated hiring tools classified as "high-risk AI systems" requiring conformity assessment, logging, and human oversight before deployment

EEOC ACLU/Aon joint submission: AI scored male candidates 8% higher than female candidates on "leadership potential" using resume language patterns alone

Neurodivergent candidates (ADHD, autism) rated 15% lower on "culture fit" by AI interview sentiment analysis compared to neurotypical peers giving identical answers

NBER Working Paper 30886: AI-assisted hiring raised overall hire rates by 7.8% across 480,948 applications — but adoption was uneven across demographic groups

UK Equality and Human Rights Commission: 1 in 5 HR managers surveyed admitted their AI tool had produced "unexpectedly biased shortlists"

Job applicants from lower-income zip codes scored 9% lower on AI job-matching algorithms even after controlling for qualifications — proxy discrimination via address

Applicants with employment gaps of 6+ months received 45% fewer AI-generated interview invitations regardless of reason for gap

Résumés from candidates at HBCUs received 20% fewer AI screening passes than identical résumés from PWIs

AI proctoring tools used in screening assessments flagged dark-skinned candidates at 3x the false-positive rate for "suspicious behavior"

Women were 79% less likely to be shown high-paying STEM job ads by Google's ad targeting algorithm in a 2015 field experiment

SHRM survey: 79% of HR professionals report using AI in at least one stage of hiring as of 2024

Only 9% of companies using AI in hiring have conducted an independent third-party bias audit of those tools

Candidates over 50 were 35% less likely to progress past AI screening stage than candidates under 35 with equivalent qualifications

AI models trained on historical hiring data perpetuate legacy discrimination: if a company historically hired 80% men, AI learns to prefer male applicant signals

UK audit: AI scoring of spoken interview responses rated regional English accents 14% lower on "professionalism" vs. Received Pronunciation

Personality screening tools used in hiring produce adverse impact scores below the 4/5ths rule for Black applicants in 62% of validated audits reviewed

Japan: AI screening tools showed 3x higher rejection rate for candidates with kanji-only names versus katakana-romanized names

AI cover letter scanners mark down candidates who use gender-neutral pronouns or disclose pronouns in signatures

FTC warning (2023): AI hiring vendors that make unfounded "predictive validity" claims about candidate success risk deceptive trade practices charges

Autistic candidates self-reported 41% lower performance on timed AI gamified assessments compared to interviews despite equivalent technical qualifications

ADHD candidates scored 18% lower on AI response-time metrics in cognitive screening games — not due to lower ability but slower initiation patterns

Candidates who self-disclosed disability in application forms were 28% less likely to receive an AI-generated interview invitation, even when disclosure was legally protected

ACLU analysis: predictive models applied to employment predict "job risk" at 2.3x higher rate for Black males than white males with same job history

Video interview AI: candidates using home backgrounds scored 7% lower on "professionalism" — disadvantaging lower-income applicants without neutral backdrop

Workers with prison records are screened out by 96% of AI applicant tracking systems before a human ever reviews their application

Candidates whose résumés mention religious organizations receive 20% fewer AI-shortlist calls — indirect discrimination on religious grounds

Workday's Responsible AI framework (2023) includes disparate impact testing across 8 protected classes — one of the first enterprise vendors to publish detailed audit methodology

Colorado SB 205 (2024): insurers and employers using AI for high-stakes decisions must demonstrate no unfair discrimination by protected class

Résumés from candidates with non-Anglo names required 27% more applications to get the same callback rate as Anglo-named candidates in AI-screened pipelines

AI language model GPT-4 when asked to rank candidates for "senior manager" roles showed 14% preference for male pronouns in generated shortlists

Only 37% of HR technology vendors disclose what training data their AI hiring tools use when asked by enterprise clients

UK CIPD survey: 55% of workers say AI-generated rejection without human explanation left them feeling treated as "less than human" in the hiring process