What Employers Should Know About AI’s Disparate Impact in Recruitment

Artificial intelligence (AI) is changing how employers recruit, screen, and evaluate candidates, as AI-enabled tools can help hiring teams manage large applicant pools, identify qualifications faster, and create more consistent workflows. Even further, employers can use these technologies to score resumes, conduct online assessments, support video interviews, or determine whether applicants meet baseline qualifications.

But this efficiency does not come risk-free. Although using AI in hiring is not automatically unlawful, these tools (while neutral on their face) can create "disparate impact" if they disproportionately exclude applicants in a protected group.  

Safety-sensitive employers should treat the use of AI hiring tools, and the bias issues that they can create, with the utmost care.

What AI Disparate Impact means in Recruitment

Disparate impact occurs when a facially neutral employment practice disproportionately excludes people in a protected group.  

Note that disparate impact is different from disparate treatment, which involves intentionally treating someone differently because of a protected characteristic. The latter requires intent; the former need not have discriminatory motivations.  

In recruitment, this distinction matters. Consider, for instance, these scenarios.

• A resume screener that prioritizes certain job titles, schools, employers, keywords, or uninterrupted work histories.  
• A video assessment that evaluates tone, eye contact, facial movement, response speed, or speech patterns.  
• A chatbot that rejects applicants who do not answer a question in a specific format.  

None of these factors may directly ask about race, sex, disability, age, or another protected characteristic. Yet each can still contribute to adverse impact if the tool disproportionately screens out qualified candidates from protected groups. For example, a resume screener could disadvantage older, otherwise-qualified applicants with nontraditional work histories; a video assessment could disadvantage applicants with speech, hearing, visual, neurological, mobility, or mental health disabilities; and a chatbot could reject applicants using screen readers or assistive technology.  

AI resume screening bias is another common example. A model trained on past hiring decisions may learn to favor candidates who resemble prior hires, and if those historical hiring patterns favor certain credentials, career paths, neighborhoods, schools, or employment histories, the tool may repeat those patterns at scale. The result can be algorithmic bias that looks data-driven but is not necessarily fair, job-related, or compliant.

What the EEOC says about AI, Title VII, and employment selection

The US Equal Employment Opportunity Commission (EEOC) has identified AI and automated systems as potential sources of employment discrimination, with most of the agency's guidance and related resources focusing on one important practical point: federal employment discrimination laws apply whenever employers use software, algorithms, or artificial intelligence in employment decisions.

For instance, while employers may use employment tests and other selection procedures under Title VII, those procedures can create risk when they disproportionately screen out, derank, unfavorably assess, or exclude applicants based on race, color, religion, sex, or national origin for reasons that are not job-related or consistent with business necessity.

Furthermore, just because an employer uses a third-party vendor, it does not eliminate the need to comply with all applicable federal hiring laws. Hiring teams need to review and understand each vendor’s validation study, audit report, or product documentation, and when the vendor in question uses AI as part of an employment selection procedure, the employer in question should be prepared to show what the tool measures, why those factors matter for the role, and how outcomes are monitored for adverse impact. 

How AI Hiring Tools May create ADA Risk

AI bias in hiring can also create risk under the ADA, if a specific hiring technology or AI tool screens out a qualified applicant with a disability, fails to support reasonable accommodation, or measures disability-related limitations instead of job-related skills.

• For example, a timed assessment may disadvantage an applicant who needs extra time as a reasonable accommodation.  
• A game-based assessment may measure reaction speed or visual processing in a way that is not essential to the job.  
• A video assessment may penalize a candidate with a speech impairment, autism, a visual disability, or a mobility-related difference even when those traits are unrelated to successful job performance.

Video assessment bias is particularly important to consider because some tools attempt to evaluate communication style, facial expression, voice, or behavior. If the tool interprets disability-related characteristics such as lower engagement, lower confidence, or lower fit, the employer may unintentionally exclude qualified candidates.

Employers should build reasonable accommodation pathways into the hiring process before using AI-enabled assessments, and applicants should know what technology will be used, what the tool is designed to evaluate, and how to request accommodation. Additionally, such a request should not reduce the candidate’s chances.

Common examples of AI-related disparate impact in recruitment

Here are some additional examples of AI-related disparate impact, which may trigger EEOC, Title VII, or ADA audits.  

Historical hiring data

Models trained on past hiring data can reproduce the assumptions, preferences, and imbalances embedded in prior decisions. If a company’s historical hiring favored certain schools, employers, locations, or career paths, an AI model may treat those patterns as indicators of success, even when they are not necessary for the role.

Proxy variables

While automated hiring tools may not use protected characteristics directly to determine outcomes, they may rely on variables that operate as proxies. ZIP code, commute distance, school attended, availability windows, employment gaps, and certain keyword patterns can sometimes correlate with protected characteristics.  

Automated rejection without human review

When automated tools reject applicants without meaningful human oversight, employers face increased federal employment law compliance risk. A qualified candidate may be screened out because of missing keywords, inaccessible assessment design, or a rigid chatbot response pattern; human review can help mitigate these risks, as well as identify when an automated result may be incomplete, inaccurate, or affected by context the tool did not consider. 

AI Hiring Risk Matrix 

AI hiring risk matrix showing where compliance risks can arise across the candidate journey. 

How employers can reduce AI hiring bias

These steps can help to reduce your business's chance of inadvertent AI hiring bias and discrimination in your hiring practices.

• Start with a clear inventory. Identify every automated hiring tool you use in your sourcing, screening, assessment, interviewing, ranking, and decision support, including tools embedded in applicant tracking systems, third-party assessments, job advertising platforms, scheduling tools, chatbots, and video interview products.
• Map where each tool affects the candidate journey. Does the tool decide who sees a job ad? Does it determine who is eligible to apply? Does it rank resumes? Does it reject applicants? Does it influence interview selection?  
• Monitor your hiring outcomes. Review your selection rates where legally permissible and appropriate, and when a tool results in adverse action, make sure the selection procedure was job-related and consistent with business necessity.

Additionally, make sure to always include humans in the loop: Your recruiters and hiring managers should understand how each AI tool your company uses works, what their limitations are, when to escalate concerns, and how to document decisions. A human reviewer should not simply accept the tool’s output without context. 

Frequently Asked Questions About AI Disparate Impact in Recruitment

Bottom line

AI can support faster, more consistent hiring, but no matter what AI tools or vendors you use, the ultimate compliance obligations remain with the employer. Do not treat AI as a one-time technology purchase; "set it and forget it," as it were. Treat your AI hiring system as an employment selection procedure that requires validation, monitoring, documentation, accommodation, and human oversight.

Know what tools are being used, understand how decisions are made, evaluate outcomes, and keep people accountable for the process. That is the path to reducing AI hiring bias while maintaining efficient, compliant hiring practices.

How DISA Can Help

DISA Global Solutions helps employers strengthen hiring, screening, and compliance practices in a changing regulatory environment. As AI becomes more common in recruitment, employers need practical workflows that support consistency, documentation, candidate communication, and compliance oversight.

DISA’s employment screening solutions help organizations build structured, role-relevant hiring programs that support informed decision-making. From background screening and pre-employment background checks to compliance-focused resources, DISA works with employers to simplify complex hiring processes while supporting safer, more responsible workforce decisions.

AI can improve recruiting efficiency, but employers still need processes that are transparent, consistent, and defensible. 
Learn how DISA helps employers strengthen hiring compliance and screening practices in a changing regulatory environment.