# AI bias in hiring

AI bias in hiring is when an automated hiring tool produces systematically skewed outcomes for candidates based on protected characteristics like race, gender, or age, usually without anyone building the tool intending that result.

It typically comes from the data a system was trained or calibrated on, not from an explicit rule targeting a group.

**AI bias in hiring** — AI bias in hiring is when an automated tool, whether it screens resumes, scores interviews, or ranks candidates, produces systematically different outcomes for one group of candidates versus another based on a protected characteristic, rather than based on actual job-relevant qualifications. It usually isn't the result of an explicit instruction to favor or exclude a group. More often it comes from the data a model was trained on reflecting historical hiring patterns that were themselves uneven, which the model then learns and repeats at scale. A well-known example is a resume-screening tool trained on a company's past successful hires learning to favor whatever traits, including irrelevant ones correlated with a protected group, happened to be common among those past hires. Because the bias is baked into patterns rather than an explicit rule, it can be difficult to detect without specifically testing outcomes across different candidate groups.

AI bias matters because an automated tool can apply a skewed pattern to every single candidate it processes, at a scale no individual human reviewer could match, which means an undetected bias compounds fast. It's also why AI hiring tools are increasingly subject to testing and disclosure requirements: the only reliable way to catch bias that isn't explicit in the tool's logic is to measure whether outcomes actually differ across groups, which is the same underlying idea behind measures like adverse impact and the four-fifths rule.

## Frequently asked questions

### Can an AI hiring tool be biased even without any explicit rule about protected groups?

Yes, and this is the most common way it happens. A model can learn to favor traits correlated with a protected characteristic from patterns in its training data, without any rule ever referencing that characteristic directly.

### How do employers check for AI bias in a hiring tool?

Common approaches include testing whether outcomes, like selection or scoring rates, differ meaningfully across candidate groups, and reviewing what data and criteria a model was trained or calibrated on for patterns that could skew results.

### Does using AI in hiring automatically reduce bias compared to a human reviewer?

Not automatically. AI can reduce certain kinds of inconsistency between individual reviewers, but it can also encode and repeat bias from its training data at a larger scale if that bias isn't specifically tested for and addressed.

## Related pages

- [Adverse impact](/glossary/adverse-impact)
- [EEOC compliance](/glossary/eeoc-compliance)
- [Browse open roles](/jobs)

## Screen with calibration built in

Intervieux's AI interview scoring runs per-dimension calibration checks designed to keep scores grounded in the actual conversation, not a pattern the model happened to learn.

Start practicing free: https://www.intervieux.ai/register · Hire with Intervieux: https://www.intervieux.ai/employers/signup