VeraHire Bias Audit Summary

Joellie AI Bias Audit of Delightify Technology’s VeraHire Hiring Assistant. Abridged summary of results.

Joellie AI was engaged to audit VeraHire’s Resume Screening and Interview Transcript Analysis tools for disparate impact on persons of any “Component 1” category, that is race/ethnicity and sex, as the minimal requirement for a bias audit under New York City Local Law 144.

The audit did not find conclusive evidence of disparate impact in either tool.

For the announcement and what the audit means, see the press release.

Scope and method

Audit date: August 25, 2026. See the full report, Summary of Bias Audit Results, for further details.
Data: Historical US candidacies from actual resumes and interview transcripts, January 12 to August 1, 2026.
Inference: Race/ethnicity and sex were inferred from names and geography.
Population: 11,725 Resume Screening candidacies and 937 Interview Transcript Analysis candidacies.
Coverage: Race/ethnicity was resolved for every record. Sex was unknown for 1,147 resume and 112 interview candidacies, which are excluded from the sex and intersectional calculations.

Employer distribution date. Each employer or employment agency must publish the date it began using VeraHire next to its link to this summary.

Resume Screening tool

By sex

Sex	N candidacies	Scoring rate	Impact ratio
Male	5,098.2	0.5019	1.0000
Female	5,479.8	0.4798	0.9559

By race and ethnicity

Race / ethnicity	N candidacies	Scoring rate	Impact ratio
White	5,422.2	0.4947	1.0000
Black or African American	1,911.0	0.4908	0.9920
Two or more races	730.1	0.4831	0.9764
Hispanic or Latino	1,608.7	0.4736	0.9573
Asian	2,042.8	0.4732	0.9565
American Indian or Alaska Native (AIAN)	9.7	0.4451*	Excluded < 2%
Native Hawaiian or Pacific Islander (NHOPI)	< 0.5†	—	Excluded < 2%

By race, ethnicity and sex

Race / ethnicity	Sex	N candidacies	Scoring rate	Impact ratio
Hispanic or Latino	Male	628.7	0.4688	0.9130
Hispanic or Latino	Female	873.0	0.4805	0.9357
Non-Hispanic White	Male	2,573.0	0.5135	1.0000
Non-Hispanic Asian	Male	908.3	0.5045	0.9826
Non-Hispanic Black or African American	Male	683.1	0.4872	0.9488
Non-Hispanic Two or more races	Male	301.2	0.4976	0.9691
Non-Hispanic AIAN	Male	3.9	0.5162*	Excluded < 2%
Non-Hispanic NHOPI	Male	< 0.5†	—	Excluded < 2%
Non-Hispanic White	Female	2,683.7	0.4810	0.9367
Non-Hispanic Asian	Female	558.2	0.4470	0.8705
Non-Hispanic Black or African American	Female	1,041.5	0.4975	0.9689
Non-Hispanic Two or more races	Female	317.9	0.4692	0.9138
Non-Hispanic AIAN	Female	5.0	0.3977*	Excluded < 2%
Non-Hispanic NHOPI	Female	< 0.5†	—	Excluded < 2%

Interview Transcript Analysis tool

By sex

Sex	N candidacies	Scoring rate	Impact ratio
Female	430.5	0.4965	1.0000
Male	394.5	0.4544	0.9151

By race and ethnicity

Race / ethnicity	N candidacies	Scoring rate	Impact ratio
Two or more races	54.2	0.5086	1.0000
Hispanic or Latino	126.8	0.4945	0.9724
Black or African American	121.8	0.4738	0.9317
White	382.6	0.4638	0.9120
Asian	250.5	0.4602	0.9050
AIAN	0.6	0.5073*	Excluded < 2%
NHOPI	< 0.5†	—	Excluded < 2%

By race, ethnicity and sex

Race / ethnicity	Sex	N candidacies	Scoring rate	Impact ratio
Hispanic or Latino	Male	50.9	0.4107	0.7441‡
Hispanic or Latino	Female	69.0	0.5519	1.0000
Non-Hispanic White	Male	170.9	0.4393	0.7960‡
Non-Hispanic Asian	Male	105.4	0.4967	0.9000
Non-Hispanic Black or African American	Male	45.0	0.4552	0.8249
Non-Hispanic Two or more races	Male	22.1	0.4674	0.8469
Non-Hispanic AIAN	Male	0.2	0.4769*	Excluded < 2%
Non-Hispanic NHOPI	Male	< 0.5†	—	Excluded < 2%
Non-Hispanic White	Female	197.8	0.4856	0.8800
Non-Hispanic Asian	Female	74.4	0.4860	0.8805
Non-Hispanic Black or African American	Female	65.1	0.4861	0.8807
Non-Hispanic Two or more races	Female	23.4	0.4991	0.9043
Non-Hispanic AIAN	Female	0.3	0.5217*	Excluded < 2%
Non-Hispanic NHOPI	Female	< 0.5†	—	Excluded < 2%

Notes

* The scoring rate for AIAN is reported for transparency but is based on too few candidacies for the rate to be statistically reliable, which is why the subgroup is excluded from the four-fifths impact ratio comparison.

† N is an estimate rather than an exact headcount. Native Hawaiian or Pacific Islander (NHOPI) and American Indian or Alaska Native (AIAN) are not identified on their own by the main inference method, and so were estimated from published Census name tables (see section 6 of the full report for further details). AIAN appears in those tables as its own category, which means it could be estimated for individual applicants and a scoring rate calculated. NHOPI does not. Every other category’s N is the sum of individual applicants’ fractional assignments, each of which can be paired with that applicant’s score. NHOPI can be estimated only as a total for the pool, with no individual estimates beneath it and so nothing to pair with scores. It is reported as a count only, with no scoring rate or impact ratio.

‡ These two categories fall below the four-fifths threshold: Hispanic or Latino Male at 0.7441 (95% CI 0.5204 to 1.0168) and White Male at 0.7960 (95% CI 0.6358 to 0.9987). Both are point estimates computed on small intersectional cells, and both confidence intervals (CI) span 0.80, so the data cannot distinguish a genuine four-fifths shortfall from sampling variation in either case.

About the auditor

Joellie AI is an independent AI governance and auditing firm that helps organizations understand, assess, and manage the risks created by artificial intelligence.

Their work spans independent AI audits and assurance, technical evaluations and AI governance consulting. Their team brings experience in independent AI auditing, AI governance, responsible AI, and emerging regulatory frameworks, with particular expertise in areas such as NYC Local Law 144, the EU AI Act, NIST AI RMF, and ISO/IEC 42001.

https://joellie.ai/

Delightify Technology · VeraHire Hiring Assistant. Questions about this audit summary: privacy@verahire.ai