How to Reduce Hiring Bias Using Applicant Tracking Systems

Your applicant tracking system is either your most powerful bias-reduction tool or your most efficient bias amplifier, and the difference comes down entirely to how it is configured and audited. Most HR teams get this wrong, not because they lack intention, but because every competing article on this topic gives them the wrong starting sequence.

Yet only 26% of job applicants trust AI will fairly evaluate them, a gap that signals a real and growing credibility problem that HR leaders must address head-on.

This guide is written for HR professionals and talent acquisition leaders who already have an ATS and are under pressure from DEI teams, legal counsel, or their own data to make it more equitable. We will walk you through the correct sequence: audit your liability first, then configure your features, then measure continuously.

Key Takeaways

• Audit first: Perform an independent bias audit prior to the deployment of your AI-enabled hiring capabilities in the ATS.
• Compliance with regulations is key: Regulations such as NYC Local Law 144, Illinois HB 3773, and Colorado SB 24-205 call for transparency, audits, and candidate notifications.
• Bias may happen in all hiring steps: Each step of the process, including job description, resume screening, AI scoring, and video interviewing, needs to be assessed frequently.
• ATS setup matters: Make use of blind resume screening, interview scorecards, skills-based hiring, and DEI metrics in order to minimize bias.
• Continuous measurement of hiring fairness: Keep track of adverse impact, diversity ratios, time-to-hire, interviewer reliability, and post-hire results in order to secure fairness.
• Vendor liability is important: Ensure that the vendor is capable of allowing for independent audits and demographic outcome data.
• Fair hiring benefits the business: Compliance and bias awareness will enhance employer brand and widen talent pool.

Is Your ATS Already a Liability?

This is the step every competing article skips. They tell you to turn on the blind screening process and add structured scorecards. But if your ATS uses AI-powered scoring or even weighted keyword filters, you may already be in violation of employment law before you configure a single new feature.

Before touching any ATS setting, your first action should be to answer three questions:

1. Does your ATS use AI scoring or automated ranking? If yes, map every feature that influences candidate ranking.
2. Are you hiring anyone who lives in New York City, Illinois, or Colorado? State-specific AEDT laws may apply even if you have no office in that state.
3. When did you last run a bias audit, and who conducted it? Vendor self-audits do not satisfy legal requirements.

The EEOC recovered almost $700 million for discrimination victims in FY2024, highlighting the legal exposure and costs associated with biased hiring practices.

Which Laws Apply to Your Right Hiring Process?

The regulatory map for AI-powered hiring is patchy but consequential. Federal enforcement has deprioritized disparate impact cases under the 2025 executive order changes, but state laws have not moved and, in some cases, have strengthened.

New York City Local Law 144 (LL 144)

This is the most technically detailed AEDT law in the US, and the one most diverse teams misunderstand. Three things you must know:

• Remote-hire trigger: LL 144 applies whenever you hire a New York City resident for any role, including fully remote. You do not need a NYC office. Violations carry $500–$1,500 per day per tool.
• Employer obligation: Compliance is your responsibility, not your vendor’s. A vendor’s internal audit does not satisfy LL 144; you need an independent third-party audit conducted annually.
• Candidate rights: You must provide candidates with a 10-day advance notice before using an AEDT tool and offer a reasonable alternative assessment on request.

Illinois HB 3773 and Colorado SB 24-205

Illinois requires written disclosure to applicants when AI is used in hiring and mandates annual demographic reporting of AI screening outcomes. Colorado’s law focuses on algorithmic discrimination standards for high-risk AI systems, including hiring tools. Both laws remain in full force regardless of federal posture. Setting diversity goals is essential as it keeps the issue of diversity front and center within organizations, encouraging accountability and focus on hiring practices.

Title VII and the Four-Fifths Rule

Adverse impact under Title VII is still measured using the four-fifths (80%) rule: if one demographic group’s selection rate is below 80% of the highest-selected group’s rate, adverse impact is indicated. But the EEOC’s 2023 technical assistance document clarified a critical point that most content gets wrong.

Where Hiring Bias Actually Lives in an ATS Workflow?

The framing that ATS systems are bias reducers is dangerously incomplete. University of Washington research (2024) found significant racial, gender, and intersectional bias in how three leading large language models ranked job applicants based solely on names. The bias is not hypothetical. It is measurable, documented, and present in the systems most organizations are running today.

Bias enters ATS workflows at four distinct stages:

Stage 1: Job Description Language

The code words ‘rock star’, ‘ninja’, and ‘dominant’ act as a statistical deterrent to women applying for these positions. A poorly phrased job posting disqualifies potential candidates before their resumes are ever scanned by the ATS. Utilize a language check like Textio or Gender Decoder as part of your standard pre-publishing process. Measuring hiring success against predetermined diversity targets will lead to an increased presence of underrepresented managers over time.
This is the configuration issue that no competing article addresses: keyword lists are created by recruiters, and recruiters have biases. You may be thinking you’re filtering for skills when you hard filter ‘Stanford’, ‘Goldman Sachs’, or ‘Fortune 500’. You are really filtering for economic class through the keyword list.

Stage 2: AI Scoring Models

AI models trained on historical hiring data inherit historical bias. Forbes (2026) cites research showing that AI hiring systems often inherit or exacerbate human biases embedded in training data because models are trained on historical hiring decisions that reflect decades of inequitable practices.

Stage 3: Video and Async Assessment Tools

AI-analyzed video interviews introduce additional bias vectors: lighting, accent, camera quality, and background environment all influence AI scoring in ways that correlate with socioeconomic status and ethnicity. If your ATS integrates a video screening tool, it requires its own independent bias audit separate from the core ATS review.

The Right Sequence: Compliance First, Configuration Second

Every major competitor article starts with feature configuration. That is the wrong order. Here is the correct sequence:

ATS Feature Toolkit: What to Configure and How

Once your compliance baseline is established, these are the features with the strongest evidence base for reducing bias at scale. Collaborative hiring reduces the impact of any single individual’s bias and prevents groupthink by requiring multiple team members to score candidates independently. Regularly auditing ATS data is critical to ensure automated tools aren’t magnifying historical human biases.

Resume Anonymization and Blind Screening

Anonymization removes the name, photo, graduation year, address, and institution from the resume display during initial screening. Franklin Electric reduced time-to-hire by 55% across 20 countries after implementing anonymized screening, demonstrating that equity and efficiency are not in tension.

Structured Evaluation Scorecards

Structured interview questions reduce bias more reliably than any other single intervention. Effective unconscious bias training does more than just raise awareness; it actually reduces bias in attitudes and behaviors at work, from hiring decisions to everyday workplace interactions.

DEI Analytics and Funnel Tracking

Tracking diversity ratios at every pipeline stage is how you find where bias is operating, not just whether it is operating. Awareness training is the first step to unraveling unconscious bias because it allows employees to recognize that everyone possesses them and to identify their own biases.

Skills-Based Matching Configuration

Blind hiring practices are most effective when combined with a structured interview stage, as this approach helps prevent bias from re-entering the process later, such as during face-to-face interviews. US states have moved toward skills-based hiring practices to expand talent pools, a signal that skills-based approaches are becoming the regulatory expectation rather than just a best practice.

Independent Bias in Hiring Audits for Applicant Tracking Systems

An independent bias audit is a structured analysis of your ATS’s demographic outcomes by a third party with no commercial relationship to your ATS vendor. This is a legal requirement under LL 144 and a best practice under Title VII, regardless of geography.

What a Compliant Audit Covers

• Analysis of the adverse impact at every automated decision point in the hiring funnel
• Four-fifths calculations with statistical significance testing
• Intersectional cross-tab analysis (race × sex at minimum)
• Documentation of training data sources and model update history for AI-scored features
• A written audit summary suitable for regulatory submission

The Vendor Data-Sharing Problem

This is a real, underreported operational risk: some ATS vendors refuse to share the demographic outcome data needed to conduct a bias audit. Before signing or renewing any ATS contract, negotiate an explicit audit-rights clause that requires the vendor to provide disaggregated demographic outcome data to any auditor you authorize.

Audit Costs and Timeline

Independent audits for mid-size ATS implementations typically range from $5,000 to $25,000, depending on the number of AEDT features, data volume, and the depth of intersectional analysis required. Budget for annual re-audits, LL 144 requires them, and best practice under Title VII supports them.

What Gets Measured with Applicant Tracking Metrics for Fair Hiring?

Reporting that your bias-reduction program is ‘in place’ is not the same as proving it is working. These are the metrics that demonstrate actual progress for unbiased hiring:

Pipeline Diversity Ratios by Stage

Track the demographic composition of your candidate pipeline at application, screen, interview, offer, and hire stages. The goal is not identical numbers at every stage; it is understanding where the funnel narrows disproportionately and why.

Time-to-Hire by Demographic Group

Disparate time-to-hire is a documented form of process bias. If candidates from certain groups consistently wait longer between stages, that gap requires investigation even if final selection rates appear balanced.

Interview Pass Rates by Interviewer

Structured scorecards generate data. Analyzing scorecard ratings by interviewer surfaces individual raters whose scores diverge significantly from panel consensus, a signal of either bias or a calibration gap that training can address.

Post-Hire Performance and Retention by Source

If your bias-reduction measures are working, you should see improved quality of hire across demographic groups over time, not just more diverse hiring. This reported improvement in post-hire performance demonstrates that equitable process and quality of hire are reinforcing, not competing, goals.

Real Insights: Implementation of AI-based Resume Screening

As per app 365 study a company in the United States used AI to enhance the applicant tracking system of hiring processes. It included using Natural Language Processing and machine learning for the analysis of candidates' resumes in respect of job-related skills and experience without taking into account any personal data such as names, gender, and age.

There is no doubt that the use of AI resulted in remarkable improvements. The firm managed to reduce the time-to-hire by 60%, increase the number of qualified candidates by 45%, and decrease the amount of work done manually to screen resumes by 35%. The fact that AI worked in the ATS made the recruiters' lives easier.

Conclusion

Debiasing through ATS does not come as an “on” or “off” switch for a particular technology. This is an approach of regulatory compliance, metrics, and change management that encompasses your entire hiring ecosystem. Those organizations that are doing this well have several things in common – namely, they audit first before they configure, treat the keyword list itself as a source of bias, conduct third-party audits instead of relying on vendor self-reviews, and integrate monitoring into their regular operating calendar. This gives them a competitive advantage in talent acquisition.
The place to begin is with the audit. Determine your risk level. And then move forward in designing the configurations, monitoring, and measurements.