The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees

Employers now have access to a wide range of computer-based tools to help with hiring, tracking employee performance, setting pay and promotions, and defining workplace policies. These tools are often used to save time, improve efficiency, increase fairness, or reduce bias. However, they may unintentionally create barriers for job applicants and employees with disabilities. When that happens, employers could be at risk of violating federal Equal Employment Opportunity (EEO) laws that protect individuals with disabilities.

This document provides answers to common questions about how software that uses algorithmic decision-making can conflict with Title I of the Americans with Disabilities Act (ADA). It also offers practical advice to employers on how to comply with the ADA and guidance for job seekers and employees who believe their rights have been violated.

The Equal Employment Opportunity Commission (EEOC) enforces federal EEO laws that prohibit workplace discrimination based on race, color, national origin, religion, and sex – including pregnancy, sexual orientation, and gender identity – as well as disability, age (40 and older), and genetic information. This publication is part of the EEOC’s ongoing effort to inform employers, workers, and other stakeholders about how EEO laws apply when businesses use employment software, including those that rely on algorithmic decision-making.

Background

This section defines three essential terms – software, algorithms, and artificial intelligence (AI) – and explains how they are connected in the workplace.

Software

Software refers to programs or procedures in information technology that provide instructions to a computer to complete a task. "Application software," also called an "application" or "app," is designed to help users carry out specific tasks. The United States Access Board provides these definitions.

Various types of software and applications are commonly used in employment, including automatic resume-screening tools, hiring platforms, chatbot software for recruitment and workflow management, video interviewing tools, analytics programs, employee monitoring software, and worker management applications.

Algorithms

An algorithm is a set of instructions that a computer follows to complete a task. In human resources software, algorithms help process data to assess, rank, and make decisions about job candidates and employees. These algorithm-driven tools are used at different points in the employment process, such as hiring, performance reviews, promotions, and termination.

Artificial Intelligence (AI)

Employers and software developers often use AI when designing algorithms that assess job applicants and employees. The National Artificial Intelligence Initiative Act of 2020 defines AI as a machine-based system that makes predictions, recommendations, or decisions based on human-defined objectives, affecting real or virtual environments. In employment, AI usually means that the software relies on data analysis to determine which factors to consider in decision-making. AI technologies can include machine learning, computer vision, natural language processing, intelligent decision support systems, and autonomous systems. The National Institute of Standards and Technology Special Publication 1270 offers further discussion on AI and bias management.

Employers may use different types of software that integrate algorithmic decision-making at various employment stages. Examples include:

• Resume-screening tools that rank applications based on specific keywords.
• Employee monitoring programs that track productivity through keystrokes or other metrics.
• Chatbots or virtual assistants that ask job-related questions and automatically reject candidates who do not meet set criteria.
• Video interviewing platforms that analyze facial expressions and speech patterns.
• Assessment software that assigns "job fit" scores based on personality, cognitive skills, or cultural compatibility through game-based or traditional tests.

ADA Basics

1. What is the ADA and How Does It Define Disability?

The Americans with Disabilities Act (ADA) is a federal civil rights law that protects individuals with disabilities from discrimination. Under Title I, the law applies to employers, employment agencies, labor organizations, and joint labor-management committees with at least 15 employees, prohibiting them from making employment decisions based on disability. Other sections of the ADA, not covered here, ensure people with disabilities have equal access to public and private services and facilities.

The ADA has a specific definition of what qualifies as a current disability. A physical or mental impairment falls under this definition if, when left untreated, it "substantially limits" one or more "major life activities." These activities include essential functions such as seeing, reaching, speaking, concentrating, or communicating, as well as the operation of major bodily systems like brain or neurological functions. (Other definitions of disability exist under the ADA, but they are not covered in this section. For more details, refer to the EEOC’s Questions and Answers on the ADA Amendments Act.)

A condition does not have to be severe, long-term, or cause extreme functional limitations to meet the standard of being substantially limiting. If it makes tasks more difficult, painful, or time-consuming compared to how most people perform them, it may still qualify. Even conditions with symptoms that come and go may be considered a disability if they substantially limit a major life activity when active. Many common medical conditions meet this standard.

2. How Can an Employer’s Use of Algorithmic Decision-Making Tools Violate the ADA?

Employers who use algorithm-based decision-making tools could violate the ADA in several ways:

• Failure to Provide Reasonable Accommodations: If an employer does not offer a reasonable accommodation needed for a job applicant or employee to be evaluated fairly by the algorithm, this could lead to discrimination. (See Questions 4–7 below.)
• Unintentional or Intentional Screening Out of Individuals with Disabilities: If an algorithm eliminates or ranks a candidate lower due to a disability, even though the individual could perform the job with reasonable accommodations, this is a violation. "Screening out" happens when a disability interferes with a selection criterion, leading to a lower score or making it impossible to complete the assessment. This could happen if the algorithm does not account for certain circumstances, reduces the accuracy of evaluations, or prevents participation altogether. (See Questions 8–12 below.)
• Violating the ADA’s Rules on Disability-Related Inquiries and Medical Examinations: Employers must ensure that algorithmic tools do not ask disability-related questions or require medical exams in ways that go against ADA regulations. (See Question 13 below.)

3. Is an Employer Responsible for Algorithmic Decision-Making Tools If They Are Provided by a Third Party?

In many cases, yes. For instance, if an employer uses a pre-employment test that discriminates against individuals with disabilities, the employer may still be responsible under the ADA, even if the test was created by an external vendor. Additionally, employers can be held accountable for the actions of third-party agents, such as software vendors, if those vendors are acting on behalf of the employer.

Algorithmic Decision-Making Tools and Reasonable Accommodations

4. What Is a Reasonable Accommodation?

A reasonable accommodation is a change in the hiring process, job duties, or workplace environment that enables a job applicant or employee with a disability to apply for a job, perform essential tasks, or have equal access to employment benefits. Accommodations may include specialized equipment, different testing methods, a quieter workspace, or exceptions to workplace rules. However, employers are not required to lower job performance standards or remove essential job functions as part of an accommodation.

5. Can Employers or Software Tools Announce the Availability of Reasonable Accommodations?

Yes. Employers can inform applicants and employees about the evaluation process and invite them to request accommodations if needed. For example, if the hiring process includes a video interview, the employer or software vendor can notify applicants about this and provide an option to request accommodations. This practice helps ensure compliance with the ADA.

6. Are Employers Required to Provide Reasonable Accommodations When Using Algorithmic Decision-Making Tools?

Yes. If an applicant or employee informs an employer that a medical condition makes it difficult to take a test or may lead to an inaccurate assessment, they are requesting a reasonable accommodation. They do not need to reference the ADA or specifically use the term "reasonable accommodation" when making this request.

Employers must respond quickly to such requests. If it is not immediately clear whether the individual has a disability covered under the ADA and requires an accommodation, the employer may request medical documentation. If the documentation confirms that a disability affects test-taking or assessment accuracy, the employer must provide an alternative testing format or a more accurate evaluation method – unless doing so would create significant difficulty or cost (known as "undue hardship").

For example, a job applicant with limited manual dexterity due to a disability may struggle with a keyboard-based test, particularly if the responses are timed. In this case, the test would not fairly measure the applicant’s knowledge. The employer would need to provide an accessible version of the test, such as one allowing verbal responses, unless this would create undue hardship. If an accessible version is not possible, the employer must consider offering a different test that assesses the same skills.

Other possible accommodations include extended time, alternative test formats, or compatibility with accessible technology, such as screen readers. Employers must ensure that applicants or employees receiving accommodations are given equal consideration to those who do not need accommodations.

Under the ADA, any medical information gathered as part of a reasonable accommodation request must be kept confidential and stored separately from an applicant’s or employee’s personnel records.

7. Is an Employer Responsible for Providing Reasonable Accommodations When Using Algorithmic Decision-Making Tools Developed by a Third Party?

In many situations, yes. As noted in Question 3, employers can be held responsible for the actions of third parties, such as software vendors, if they have authorized these entities to act on their behalf. For instance, if an employer hires a vendor to create and administer a pre-employment test, the employer may still be liable for any discrimination caused by the vendor’s actions or inactions. If an applicant informs the vendor that a medical condition is making the test difficult (which qualifies as a request for reasonable accommodation), and the vendor fails to provide the necessary accommodation required under the ADA, the employer could still be responsible – even if it was unaware of the issue.

Algorithmic Decision-Making Tools That Exclude Qualified Individuals with Disabilities

8. What Does It Mean to Be Screened Out Due to a Disability, and When Is It Unlawful?

Screening out happens when a disability prevents a job applicant or employee from meeting a specific requirement or lowers their assessment score, leading to a lost job opportunity. Under the ADA, this type of exclusion is unlawful if the individual can still perform essential job functions with reasonable accommodations, provided that accommodations are legally required.

Questions 9 and 10 explain how screening out occurs, and Question 11 provides examples of when an individual excluded due to a disability may still be qualified for the job with accommodations.

9. Can Algorithmic Decision-Making Tools Exclude Individuals with Disabilities? What Are Some Examples?

Yes. If an algorithm causes an individual to receive a lower score due to a disability, and that lower score results in a lost job opportunity, the tool has effectively screened out the individual based on their disability.

One example involves chatbots – automated software that communicates with applicants through text, email, or online interactions. A chatbot might be programmed to automatically reject candidates who mention significant employment gaps. If an applicant has a gap in work history due to a disability, such as taking time off for medical treatment, the chatbot could unintentionally exclude them based on that disability.

Another example is video interviewing software that evaluates speech patterns to determine problem-solving ability. If an applicant has a speech impediment, the software may score them unfairly because their speech differs from typical patterns. If this results in rejection, the individual has been screened out due to their disability.

10. If an Algorithmic Decision-Making Tool Is Marketed as Bias-Free, Does That Mean It Will Not Exclude Individuals with Disabilities?

Not necessarily. Employers and software vendors that claim their tools are bias-free usually mean they have taken steps to prevent a type of discrimination known as adverse impact or disparate impact under Title VII of the Civil Rights Act. This type of discrimination occurs when an employment policy negatively affects a particular group based on race, sex, national origin, color, or religion.

To reduce the risk of disparate impact discrimination, employers or vendors may test their tools on different demographic groups and compare the average results. If the results for one group are consistently less favorable than for others – such as lower average scores for a particular racial group – the tool may be adjusted to eliminate the disparity.

However, preventing this kind of bias is different from addressing disability bias. Since disabilities vary widely, an individual may still receive a low score due to their specific condition, even if other individuals with disabilities do not experience the same issue. To prevent discrimination against people with disabilities, employers may need to take additional steps beyond those used to minimize other forms of bias. (See Question 12.)

11. When Is Screening Out Due to a Disability Unlawful, and Can an Individual Still Perform the Job?

Screening out a candidate because of a disability is illegal if the person can perform the essential job functions with a reasonable accommodation when legally required.

In some cases, an individual excluded by an algorithmic decision-making tool may still be capable of performing the job. For instance, some employers use gamified assessments – video game-based tests – to measure memory, personality traits, or other abilities. If an employer requires a 90% score on a memory-based game, a blind applicant might be unable to complete the test and would receive a lower score. However, despite being screened out, that individual could still have an excellent memory and be fully capable of handling a job requiring strong recall skills.

Even if a decision-making tool has been validated – meaning it meets professional standards for accurately assessing job-related skills – it may still be inaccurate for certain individuals with disabilities. A gamified memory test, for example, may work well for most applicants but still unfairly screen out those who are blind simply because they cannot interact with the test’s interface.

Additionally, algorithmic tools often evaluate candidates based on typical workplace conditions without considering the reasonable accommodations available to employees with disabilities. Some personality assessments, for example, might prioritize traits found in a company’s most successful employees. A candidate with PTSD might score poorly on a test measuring attention to detail due to difficulty filtering out distractions. However, if they were provided with accommodations like a quiet workspace or noise-canceling headphones, they could perform just as well as other employees. If the algorithm screens out such candidates without considering potential accommodations, it may violate the ADA.

Some individuals who score poorly due to a medical condition might not even need an accommodation to perform their job well.

12. How Can Employers Reduce the Risk of Screening Out Qualified Individuals with Disabilities?

Employers can take several steps to minimize the risk that an algorithmic decision-making tool unfairly excludes individuals with disabilities.

Before Selecting a Tool from a Vendor:

Employers should ask vendors whether their tools were designed with accessibility in mind. Some key questions include:

• If the tool has a user interface, was it made accessible to people with a range of disabilities?
• Are job-related materials available in alternative formats? If so, which ones? Are there any disabilities the tool does not accommodate, requiring the employer to step in?
• Did the vendor evaluate whether the tool disadvantages individuals with disabilities? For example, did they assess whether the traits being measured correlate with certain disabilities?

If an employer is developing its own tool, these same considerations should be factored into the design process. Engaging experts in various disabilities can help identify potential barriers. For example, when creating pre-employment tests that measure personality or cognitive traits, working with psychologists – including neurocognitive specialists – can help prevent unintended bias against individuals with autism or intellectual, cognitive, or mental health-related disabilities.

During Implementation and Use:

Regardless of whether the employer or a vendor developed the tool, additional steps can help prevent discrimination:

• Clearly state that reasonable accommodations, including alternative tests or formats, are available for individuals with disabilities.
• Provide clear instructions on how to request accommodations.
• Give applicants and employees as much information as possible about the assessment before they take it, including what traits the tool measures, how it evaluates them, and whether certain disabilities could affect the results.

By taking these steps, individuals with disabilities can determine whether they need an accommodation. For example, an employer using an algorithm to track productivity may measure keystrokes per minute. If employees are not informed of this, someone who is blind and uses voice recognition software instead of a keyboard might receive a low productivity score, impacting their chances of promotion. If the employer informs employees of the assessment method in advance, the individual could request an alternative way to measure productivity.

Avoiding Unlawful Screen Outs:

Employers can further reduce ADA violations by ensuring that algorithmic tools only screen out candidates who are truly unable to perform essential job functions, even with reasonable accommodations. A good practice is to choose tools that measure only the abilities required for the job – accounting for accommodations that may be available.

For example, an employer hiring cashiers should ensure that a chatbot does not automatically reject candidates who cannot stand for long periods. Otherwise, an applicant who uses a wheelchair might be excluded, even though they could perform the job with a lowered register as an accommodation.

Another effective measure is to avoid algorithmic tools that assess abilities indirectly by making assumptions based on correlated traits. For instance, if a job requires strong writing skills, an employer should avoid an algorithm that assesses writing ability by analyzing an applicant’s personality traits rather than directly measuring their writing skills. This helps prevent rejecting candidates who are fully qualified but have a disability that affects personality traits unrelated to job performance.

Algorithmic Decision-Making Tools, and Disability-Related Inquirie,s and Medical Examinations

13. How Can an Employer’s Use of Algorithmic Decision-Making Tools Violate the ADA’s Rules on Disability-Related Inquiries and Medical Exams?

An employer may violate the ADA if an algorithmic decision-making tool asks disability-related questions or functions as a medical exam before a candidate has received a conditional job offer. This remains true even if the individual does not have a disability.

A tool is considered to be making disability-related inquiries if it asks applicants or employees questions likely to reveal a disability or directly asks whether they have a disability. A test qualifies as a medical exam if it seeks information about an individual’s physical or mental health or impairments.

If an algorithm can identify an applicant’s medical conditions and is used before a conditional offer is made, it would violate ADA restrictions. However, not all tools that ask health-related questions meet this definition. For example, a personality test that asks if a person is generally optimistic does not directly inquire about a disability, even if optimism may be associated with certain mental health conditions.

Still, even if a tool does not violate the ADA’s rules on disability-related inquiries or medical exams, it may still be unlawful under other ADA provisions. For instance, if a personality test assesses optimism and an individual with Major Depressive Disorder (MDD) scores poorly, leading to a lost job opportunity, this could constitute illegal screening out based on disability. As discussed in Questions 8–11, such screening is unlawful if the individual can perform the job’s essential functions with or without reasonable accommodation.

Once employment begins, an employer may only make disability-related inquiries or require medical exams if they are legally justified under the ADA.

Best Practices for Employers

14. How Can Employers Ensure Compliance with the ADA When Using Algorithmic Decision-Making Tools?

As outlined in Questions 4–7, employers must provide reasonable accommodations when legally required. Best practices to help meet this requirement include:

• Training staff to quickly recognize and process requests for reasonable accommodations, including requests for alternative test formats or reassessments after a poor result.
• Training staff to develop or provide alternative ways to evaluate candidates and employees if the current method is inaccessible or disadvantages individuals who need accommodations.
• If a third party administers the tool on the employer’s behalf, ensuring they promptly forward all accommodation requests to the employer for proper handling. Alternatively, employers can establish agreements requiring the third party to provide accommodations in line with ADA obligations.

Employers should also minimize the risk of algorithmic tools disadvantaging individuals with disabilities, whether intentionally or unintentionally. Best practices include:

• Using tools designed to be as accessible as possible to a wide range of disabilities, reducing the likelihood of unfair disadvantages. Conducting user testing can help identify accessibility issues.
• Clearly informing all applicants and employees that reasonable accommodations are available and providing clear, accessible instructions for making accommodation requests.
• Explaining in simple language – and in accessible formats – what traits the algorithm measures, how it evaluates them, and what factors might influence ratings.

Employers should also take steps to ensure algorithmic tools do not unfairly assign poor scores to individuals who can perform the job’s essential functions with reasonable accommodation. Best practices include:

• Ensuring that the tool only assesses abilities or qualifications that are truly necessary for the job – even for individuals entitled to reasonable accommodations.
• Measuring required skills and qualifications directly rather than making indirect assessments based on unrelated characteristics or scores.
• Before purchasing an algorithmic tool, confirming with the vendor that it does not ask applicants or employees questions likely to reveal a disability or seek health-related information unless the inquiries relate to a reasonable accommodation request. (The ADA allows employers to request reasonable medical documentation for accommodation requests received before a conditional job offer, when necessary, to facilitate the application process.)

Best Practices for Job Applicants and Employees Being Evaluated by Algorithmic Decision-Making Tools

15. How Can I Ensure I Am Assessed Fairly by Algorithmic Decision-Making Tools?

If you have a medical condition that could qualify as a disability under the ADA and might negatively impact your evaluation results, you should consider asking the employer about their use of these tools. This can help determine whether the tool could create challenges related to your condition. If you identify potential issues, you may request a reasonable accommodation to ensure you are assessed fairly and on equal footing with other candidates or employees.

For instance, if the hiring process includes a test, you could ask for an accessible version or an alternative test that evaluates your skills without being affected by your disability. To request an accommodation, inform a company representative – such as someone in Human Resources – or, if the employer is using a third-party vendor, notify either the vendor or the employer. You should explain that you have a medical condition and need adjustments to ensure an accurate assessment of your abilities.

If your disability and need for accommodation are not apparent or previously documented, the employer may ask for medical documentation to support your request.

If you only realize that an algorithmic tool is causing difficulties due to your disability after testing has begun, notify the employer or software vendor as soon as possible. Ask to be assessed in a way that accurately reflects your qualifications and ability to perform the job, with a reasonable accommodation if required by law.

If you receive a low score from an algorithmic tool, consider whether your health condition might have impacted your performance. For example, did your disability make it harder to complete the assessment or affect your results? If so, contact the employer or vendor immediately to explain the issue. You can request to be reevaluated using a different method or provide information on how you can effectively perform the job despite the test results.

16. What Should I Do If I Believe My Rights Have Been Violated?

If you think your rights under the ADA have been violated, the EEOC can help you explore your options. For example, if an employer or software vendor refuses to accommodate your request for a fair assessment and you believe you could do the job with reasonable accommodations, you may consider filing a discrimination charge with the EEOC.

A discrimination charge is a formal complaint stating that an employer engaged in discriminatory practices. By filing a charge, you are asking the EEOC to investigate the issue and help find a resolution under equal employment opportunity (EEO) laws.

Once a charge is filed, the EEOC will review the complaint and may launch an investigation. Mediation, an informal and confidential process where a neutral mediator helps resolve disputes, may also be available. Since a charge must be filed within 180 days of the incident (or up to 300 days if state or local anti-discrimination laws also apply), it is best to act as soon as possible. Employers are legally prohibited from retaliating against individuals who file a charge or reach out to the EEOC for assistance.

To start the process, visit the EEOC’s Online Public Portal, go to your local EEOC office or reach out by phone.