Research Pillar · Updated May 2026

AI in Hiring

What algorithmic hiring tools actually do, what the evidence says about bias and accuracy, which legal cases have been filed, and what candidates can do about it. Primary-source statistics only — no viral myths.

Compiled by JobCannon Research · 70+ verified statistics · Full stats hub →

51%

of US companies use AI in hiring (ResumeBuilder, n=948, Oct 2024)

200M+

class members in Mobley v. Workday — first AI hiring class action certified (Feb 2025)

+7.8%

hire-rate lift from AI writing assistance, lower-skilled workers (MIT/NBER, n=480,948, 2023)

Where AI enters the hiring funnel

AI tools touch hiring at five distinct stages — and each carries a different evidence base, legal exposure, and candidate implication.

1
Résumé screening
The highest-volume AI application. ATS platforms (Workday, Greenhouse, Lever, iCIMS) filter résumés before a human reads them — but not by the "75% auto-rejection" figure that circulates online. That claim has no primary source. What the evidence does show: resume-keyword mismatch and keyword stuffing both hurt, and AI models trained on historical hire data inherit historical bias.
2
Video interview analysis
HireVue, SparkHire, and competitors transcribe, tone-analyse, and sometimes flag body language. HireVue dropped facial-expression scoring in January 2021 after disability-bias criticism. The EEOC's 2022 technical guidance names video-AI explicitly as a potential ADA and Title VII risk.
3
Gamified cognitive assessments
Platforms like Pymetrics, Aon's CEB SHL, and Arctic Shores use game-based tasks to infer cognitive ability and personality. The ACLU filed an EEOC charge against Aon in December 2023 — the first federal class-wide neurodivergent AI-hiring complaint — alleging gamified assessments discriminate against autistic and Black applicants.
4
Predictive ranking and scoring
Some platforms score candidates for "culture fit" or "flight risk" using historical employee data. This is where the bias evidence is most severe: a University of Washington / AIES study (Wilson & Caliskan, 2024, NIST-funded, 3M+ résumé-job comparisons) found AI models prefer white-associated names 85% of the time vs 9% for Black-associated names.
5
Offer and onboarding prediction
The least common AI application, but growing. Predictive attrition models flag candidates as flight risks before hiring. The same bias pipelines that affect résumé screening can affect these scores — and the candidate never sees them.

What the bias evidence actually shows

Bias in AI hiring is not hypothetical — it is documented in peer-reviewed research, government enforcement actions, and class-action litigation.

Race bias

Bertrand & Mullainathan (AER, 2004, n≈5,000) found white-sounding names received 50% more callbacks. Quillian et al. (PNAS, 2017) meta-analysed 24 field experiments and found a 36% white callback advantage with no improvement over 25 years.

Gender bias

Goldin & Rouse (AER, 2000) found blind orchestra auditions increased female advance rates by 50 percentage points. Kleven et al. (2019) document a 20-30% earnings penalty for mothers vs childless women ("child penalty"). LinkedIn Economic Graph (2023) found women apply to 26% fewer jobs but are 16% more likely to get hired when they do apply.

Age bias

Neumark et al. (NBER, 2019, n=40,000+) found candidates aged 64-66 received 35% fewer callbacks than those aged 29-31. NBER WP 28379 (2020) reviewed 20+ audit studies and found consistent 20-35% age penalties.

Neurodivergent bias

22% of autistic adults in the UK are employed vs 77% who want to work (National Autistic Society, 2021). 85% of college-educated autistic adults are underemployed (Autism Speaks, 2017). The ACLU v. Aon EEOC charge (Dec 2023) is the first federal class-wide neurodivergent AI-hiring complaint.

Full citation index: AI Résumé Statistics 2026

Legal exposure: enforcement is live

Three enforcement signals matter for candidates and employers in 2026:

→EEOC/DOJ joint guidance (May 2022) explicitly names résumé-screening AI, video-interview AI, and gamified assessments as potential ADA and Title VII violations. Employers are liable even when bias is unintentional.
→EEOC v. iTutorGroup (2023) — $365,000 settlement over an AI hiring tool that screened out applicants over 40 and women returning from career breaks. First EEOC enforcement action specifically targeting AI in hiring.
→Mobley v. Workday (class certified Feb 2025) — 200M+ class members alleging Workday's screening tools discriminate by race, age, and disability. First AI-hiring class action to survive certification. Full breakdown →

Myths we have debunked — with primary sources

Misinformation about AI hiring is widespread. These articles trace each viral claim to its origin and assess what the evidence actually says.

The 75% ATS Rejection Myth — Debunked

Why the "75% of résumés rejected before a human reads them" claim has no primary source — and what actually happens at scale.

Cornell Says 20-60% of Résumés Are Filtered By ATS — Does It?

No Cornell study says this. The claim is a viral hallucination. We traced it to its origin and found nothing.

Originality.ai Claims 99% Accuracy — Why It's a Myth

Stanford HAI puts AI-detector false-positive rates above 20% for non-native English writers. Scribbr's August 2024 head-to-head shows Originality at 76% detection — not 99%.

Mobley v. Workday Explained — 200M+ Class, Feb 2025

The first class-action lawsuit certified against an AI hiring platform. What it means for employers and candidates.

How to Write a Résumé With AI Without Getting Rejected

Evidence-backed rules for using ChatGPT and Co-pilot to improve — not generate — your résumé. Includes the 7 signals that flag AI-generated content to human reviewers.

What candidates can do

The evidence points to four things that actually work — not folklore.

Apply selectively

LinkedIn Economic Graph (2023) found women who apply to fewer, more targeted roles are hired 16% more often than those who spray applications. The same logic applies universally: AI screens trained on role-fit reward relevance signals, not volume.

Understand your personality-to-role fit

AI screens calibrated for role-fit look for narrative coherence between your experience and the role requirements. Knowing your RIASEC code or Big Five profile gives you the vocabulary to articulate fit authentically — which passes both AI and human review better than keyword stuffing.

Use AI assistance to edit, not to write

MIT/NBER (2023) found AI writing assistance raised hire probability by 7.8% when used to improve candidate-written text. Resume.io (n=3,000) finds 49% of hiring managers auto-dismiss résumés they identify as wholly AI-generated. The distinction is editing vs generation.

Document bias if you encounter it

If you receive an automated rejection that you believe reflects bias, you have grounds to file an EEOC charge (US), submit a Subject Access Request (UK), or request algorithmic review under the EU AI Act (EU). Employers are required to audit high-risk AI systems under EU AI Act Article 9, effective 2025.

Frequently asked questions

How many companies use AI in hiring?

ResumeBuilder's October 2024 survey of 948 US business leaders found 51% of companies already use AI in hiring, and 82% of those apply it to résumé review. Capterra's 2024 global survey (n=3,256, 11 countries) puts HR-AI adoption at 55% worldwide.

Can AI hiring tools discriminate illegally?

Yes. The EEOC and DOJ issued joint guidance in May 2022 warning that algorithmic hiring tools can violate the ADA and Title VII even when bias is unintentional. Mobley v. Workday (class certified February 2025, 200M+ class members) and EEOC v. iTutorGroup ($365K settlement, 2023) confirm enforcement is live.

Does AI reject 75% of résumés before a human sees them?

No — this figure is a myth traced to a 2012 Preptel sales presentation; the company closed in 2013 and published no methodology. Jobscan (2024, n=3,500) and Enhancv (2024) both independently found over 92% of surveyed job seekers denied experiencing this. Large companies use ATS for workflow, not mass auto-rejection.

Does AI help candidates too, or only employers?

Both. An MIT/NBER 2023 RCT (Brynjolfsson et al., n=480,948) found AI writing assistance raised hire probability by 7.8% for lower-skilled workers and reduced wage inequality. A Harvard Business School / BCG study (Dell'Acqua et al., 2024, n=758 consultants) found GPT-4 assistance raised task completion by 12%, speed by 25%, and quality by 40%.

How can personality tests help in an AI-screened hiring process?

Personality assessments help you identify which roles match your traits so you can apply selectively — which raises your hit rate on AI screens calibrated for role-fit. Knowing your RIASEC code, Big Five profile, or Enneagram type gives you language to articulate fit authentically rather than generically, which passes both AI and human review.

Know your personality — navigate AI hiring more deliberately

AI hiring tools increasingly sort by role-fit signals. A personality assessment gives you a vocabulary for articulating that fit — and helps you identify which roles to target before the AI screen, not after.

Career Match Assessment →Big Five Personality Test