Blog

For millions of professionals, applying for a job now begins with a familiar ritual. A candidate finds an opening, updates a résumé, completes an online application and waits. Sometimes a rejection arrives within hours. In other cases, there is no response at all. The applicant may never know whether a recruiter reviewed the résumé, an automated system ranked it too low or a knockout question removed it from consideration.

This uncertainty is especially important for Hispanic professionals. Automated hiring systems promise to make recruiting faster, more consistent and less dependent on human judgment. Emerging research, however, suggests that some of these technologies can reproduce the same racial and ethnic disparities they were expected to reduce. An algorithm does not need to recognize someone explicitly as Hispanic to treat that person differently. Names, language patterns, education, geography, professional affiliations and employment history can function as proxies for ethnicity or national origin.

The evidence does not support the sweeping conclusion that every applicant tracking system discriminates against Hispanic candidates. Many applicant tracking systems simply store applications and help recruiters organize information. The greater concern involves automated résumé ranking, candidate matching, personality assessments, video analysis and artificial intelligence tools that recommend who should advance.

The emerging question is no longer whether technology can introduce bias into hiring. Research indicates that it can. The more urgent question is how Hispanic professionals can make certain their experience, potential and value are evaluated by people rather than reduced to a score they will never see.

The Difference Between an ATS and an Automated Hiring Decision

The term “applicant tracking system” is frequently used to describe almost every piece of technology involved in recruiting. In reality, these systems have varying levels of influence. A basic ATS may collect résumés, record candidate information and allow a recruiter to search for specific skills. More sophisticated platforms can parse résumés, compare applicants with job descriptions, administer assessments, assign compatibility scores and recommend which candidates should receive interviews.

That distinction matters. Software that stores a résumé is not necessarily making a consequential employment decision. Software that ranks one candidate above another, recommends rejection or prevents an application from reaching a recruiter is participating directly in the selection process.

Harvard Business School and Accenture identified approximately 27 million “hidden workers” in the United States who wanted to work more but were frequently excluded from consideration. The group included caregivers, veterans, people with disabilities, older workers, immigrants, individuals with employment gaps and people without traditional credentials. The researchers found that employers’ automated systems often relied on rigid criteria that screened out candidates who could perform the work.

For Hispanic applicants, those filters can intersect with immigration history, international education, bilingual experience, nontraditional career paths and occupational concentration. A candidate may possess the required ability but use a different job title than the system expects. Another may have completed education outside the United States. Someone who left the workforce to care for family may be penalized for an employment gap even when that gap says nothing about the person’s ability to succeed.

What the Research Says About Hispanic Applicants

One of the most directly relevant studies examined whether large language models made different hiring decisions based on names associated with race, ethnicity and gender. Researchers changed applicant names while maintaining substantially comparable hiring scenarios. They found that, under many experimental conditions, the models were more likely to favor White applicants over Hispanic applicants. In aggregate, masculine White names received the highest acceptance rates, while masculine Hispanic names received the lowest.

The results varied when researchers changed the prompts and decision environments. That variability is significant. It means the systems did not demonstrate one universal and predictable form of discrimination, but it also means that small changes in how an employer configures or instructs a model could affect the treatment of Hispanic candidates.

Another audit of GPT-based hiring applications found that artificial intelligence could generate stereotypical information when producing résumés for candidates associated with different racial and ethnic groups. Résumés created for Hispanic candidates were more likely to contain what researchers described as immigrant markers, including non-U.S. education, international work experience and indications of non-native English proficiency.

This creates a potentially troubling cycle. An AI system may associate Hispanic identity with stereotypical characteristics and then another automated system may downgrade the applicant because of those same characteristics. The discrimination does not need to appear as an explicit command to reject Hispanic candidates. It can emerge through assumptions embedded in training data, proxy characteristics and the weight assigned to particular résumé details.

A separate study simulating automated résumé retrieval across nine occupations used more than 500 résumés and 500 job descriptions. Researchers found that the models significantly favored White-associated names in 85.1% of the conditions they examined. That study documented especially severe disadvantages affecting Black men and did not provide a definitive Hispanic rejection rate. Nevertheless, it demonstrated that a name alone can influence how a résumé-screening model retrieves and ranks otherwise relevant candidates.

A 2026 study expanded the concern from isolated decisions to the structure of the hiring market. Researchers evaluated approximately 3 million applicants who submitted 4 million applications processed by algorithms from the same vendor. They found significant racial disparities and discovered that some applicants received highly consistent rejection recommendations across different jobs.

The study’s publicly reported racial findings focused on Black and Asian applicants rather than quantifying outcomes for Hispanic applicants. Its broader warning applies across demographic groups: when many employers use similar technology from the same vendors, a biased screening pattern can follow a candidate from one company to another. What appears to be a series of independent rejections may actually be the repeated judgment of a shared algorithm.

How an Algorithm Can Infer What an Applicant Never Disclosed

Most employers do not instruct their systems to search for Hispanic applicants, and race and ethnicity information collected for equal employment reporting is ordinarily separated from the hiring decision. That does not mean ethnicity disappears from the application.

A surname may be associated with Hispanic heritage. Fluency in Spanish may appear in the skills section. Membership in a Latino professional organization may be listed under leadership experience. A résumé may include employment or education in Mexico, Puerto Rico, Colombia or another Latin American country. An address or ZIP code may correlate with the demographic composition of a community.

Even when a system does not use those characteristics directly, it can learn from historical hiring decisions. If an employer’s past workforce was not representative, a model trained to identify candidates who resemble previously successful hires may treat historical exclusion as a formula for future success. The algorithm can replicate a preference without understanding its social or legal meaning.

Language is another potential source of inequality. Résumé-screening tools may reward vocabulary that appears frequently in conventional corporate career paths while undervaluing equivalent experience described differently. Bilingual professionals, first-generation college graduates, entrepreneurs, immigrants and applicants moving from smaller organizations may not use the exact phrases that appear in an employer’s historical data. The system may interpret difference as a lack of qualification.

Bias Can Be Scaled Faster Than Opportunity

Human recruiters can make biased decisions, but automation changes the scale. One recruiter may overlook several candidates. A widely deployed screening model can influence thousands or millions of applications before anyone identifies a pattern.

Automated rejection can also conceal where exclusion occurred. A candidate who interviews with a person can often evaluate the interaction and ask for feedback. Someone rejected by a ranking system may never learn which qualification was supposedly missing, whether the résumé was parsed correctly or whether a proxy for ethnicity affected the score.

This opacity makes discrimination difficult to prove. Commercial hiring algorithms are generally proprietary, employers frequently depend on vendors and candidates rarely have access to selection data. Researchers also face the challenge of measuring Hispanic identity, a broad ethnicity that includes people of different races, national origins, immigration histories and naming traditions.

Consequently, the responsible conclusion is not that all ATS platforms discriminate against Hispanic applicants. It is that credible research has established the capacity for automated hiring systems to produce racial and ethnic disparities, while Hispanic-specific outcomes remain understudied and insufficiently transparent.

The Law Still Applies When Software Is Involved

Title VII of the Civil Rights Act prohibits employment discrimination based on race, color, religion, sex and national origin. An employer does not necessarily escape responsibility because a vendor’s technology helped produce the decision.

The Equal Employment Opportunity Commission has warned that employment protections apply to software, algorithms and artificial intelligence used in recruiting, monitoring, promotion and termination. A selection process can create legal concerns even without an openly discriminatory instruction if it disproportionately excludes a protected group and the employer cannot demonstrate that the process is job related and consistent with business necessity.

The Uniform Guidelines on Employee Selection Procedures also provide a framework for examining adverse impact. Employers should therefore evaluate results, not merely accept a vendor’s promise that its technology is neutral. A system can exclude demographic groups without ever displaying explicitly racist or ethnically discriminatory language.

Responsible employers should know what their hiring systems measure, test selection rates across demographic groups, examine proxy variables, provide meaningful human review and create a process for candidates to request accommodations or challenge inaccurate results. Technology should support professional judgment, not give organizations permission to stop exercising it.

In Person Networking Can Be the Greatest Equalizer

Automated hiring is one reason in person networking has become more valuable, not less. An application asks a system to interpret a résumé. A conversation allows a person to understand the professional behind it.

Being in the room gives candidates an opportunity to communicate qualities that are difficult to reduce to keywords: confidence, judgment, curiosity, emotional intelligence, industry knowledge, resilience and the ability to establish trust. It can transform an applicant from an anonymous record into someone a recruiter, executive or professional contact remembers.

This does not mean networking eliminates discrimination. Access to influential rooms is not distributed equally, and relationships can reinforce exclusion when professional circles remain closed. Networking becomes an equalizer only when organizations intentionally create accessible spaces where professionals from different backgrounds can meet decision-makers, exchange information and build authentic relationships.

Nevertheless, human connection can create paths around the most rigid digital filters. A professional contact can recommend that a hiring manager review a résumé personally. An employee can explain why a candidate’s unconventional experience is relevant. A recruiter who has already met an applicant may recognize the name when it appears in the system. A business leader may identify potential that an automated score failed to capture.

That is the practical power of advocacy. The goal is not merely to collect contacts. It is to develop enough professional credibility that someone is willing to say, “I met this person. You should speak with them.”

Weak Ties Can Produce Strong Career Results

Networking research provides compelling evidence for the value of relationships beyond a person’s closest circle. A large-scale study published in Science analyzed experiments involving more than 20 million LinkedIn users over five years, approximately 2 billion new connections and 600,000 new jobs. The researchers found causal evidence supporting the “strength of weak ties,” the long-standing theory that acquaintances can provide access to opportunities and information that may not exist within our closest networks.

Close friends frequently know many of the same people and possess similar information. A new acquaintance from another company, profession or industry can connect someone to an entirely different opportunity. That is why entering a professional event and speaking only with people you already know can feel comfortable while limiting the potential value of the room.

For Hispanic professionals, the importance of weak ties can be especially significant. A conversation with an executive, recruiter, entrepreneur, community leader or fellow attendee may create access to a network that an applicant could not reach through an online submission. The immediate result may not be a job offer. It could be an introduction, informational conversation, referral or piece of market intelligence that changes the direction of a career.

The most valuable person in the room may be someone you have not met yet.

Use Technology, but Do Not Depend on It Exclusively

Professionals should continue applying online. They should also optimize résumés for readability, use conventional section headings, mirror legitimate terminology from job descriptions and clearly identify measurable accomplishments. A well-structured application remains necessary in a technology-driven hiring market.

It should not be the entire strategy. Applying online without developing relationships leaves too much of the decision to systems candidates cannot see, question or influence. The strongest approach combines digital preparation with human visibility.

Before attending an event, professionals should identify the types of people they want to meet and prepare a concise introduction explaining what they do, what they are seeking and what value they offer. During the event, they should resist spending the entire evening with familiar contacts. Afterward, they should follow up with specific, personalized messages and continue the relationship before asking for assistance.

Networking works because trust develops through repeated, credible interaction. One conversation can open a door, but consistent follow-up turns recognition into professional capital.

Being in the Room Is a Career Strategy

The rise of automated hiring has not made relationships obsolete. It has made relationships a form of protection against professional invisibility.

An algorithm may see an employment gap. A person can understand the caregiving responsibility, entrepreneurial experience or personal challenge behind it. A résumé parser may not recognize that two job titles represent comparable work. An industry professional can make the connection immediately. A ranking system may view an international degree as unfamiliar. A human advocate can recognize the expertise it represents.

Hispanic professionals should not have to network around discriminatory technology to receive fair consideration. Employers have the responsibility to test their systems, investigate disparities and ensure that every selection criterion measures something genuinely connected to job performance. Until greater transparency and accountability exist, however, professionals must use every available channel to make their value visible.

Being in the room cannot guarantee an opportunity. It can do something that an online application rarely accomplishes: give someone a reason to look beyond the score, recognize the person and become an advocate.

In an employment market increasingly shaped by machines, authentic human connection may be the greatest equalizer we have.

Sources

• An, H., Acquaye, C., Wang, C., Li, Z., & Rudinger, R. (2024). Do large language models discriminate in hiring decisions on the basis of race, ethnicity, and gender? arXiv.
• Armstrong, L., Liu, A., MacNeil, S., & Metaxa, D. (2024). The silicon ceiling: Auditing GPT’s race and gender biases in hiring. arXiv.
• Bogen, M., & Rieke, A. (2018). Help wanted: An examination of hiring algorithms, equity, and bias. Upturn.
• Bommasani, R., Bana, S. H., Creel, K. A., Jurafsky, D., & Liang, P. (2026). Algorithmic monocultures in hiring. arXiv.
• Gaebler, J. D., Goel, S., Huq, A. Z., & Tambe, P. (2024). Auditing the use of language models to guide hiring decisions. arXiv.
• Fuller, J. B., Raman, M., Sage-Gavin, E., & Hines, K. (2021). Hidden workers: Untapped talent. Harvard Business School Project on Managing the Future of Work and Accenture.
• National Institute of Standards and Technology. (2022). Towards a standard for identifying and managing bias in artificial intelligence. U.S. Department of Commerce.
• Rajkumar, K., Saint-Jacques, G., Bojinov, I., Brynjolfsson, E., & Aral, S. (2022). A causal test of the strength of weak ties. Science, 377(6612), 1304–1310.
• U.S. Equal Employment Opportunity Commission. (1978). Uniform guidelines on employee selection procedures. U.S. Government.
• Wilson, K., & Caliskan, A. (2024). Gender, race, and intersectional bias in resume screening via language model retrieval. arXiv