Nearly all large companies now use some form of AI in hiring, and the share of HR teams specifically using AI for resume screening has risen sharply over the past two years. That scale is exactly why the bias question matters — a biased screening step doesn't affect one hiring decision, it affects every application that passes through it.
What the research actually found
This isn't speculation. A large-scale study testing LLM-based resume ranking across tens of thousands of paired comparisons found a stark preference for names associated with white candidates over names associated with Black candidates. A separate 2025 study found bias favoring older male candidates in certain ranking scenarios. These aren't edge cases in obscure tools — they're documented patterns in the kind of general-purpose AI models increasingly used to pre-screen and rank candidates.
What regulators are doing about it
- —New York City's Local Law 144 requires companies using automated employment decision tools to run annual bias audits and publicly disclose the results
- —The EU AI Act classifies hiring and screening AI as "high-risk," with transparency obligations already enforceable — though the fuller compliance regime (formal risk assessments, bias testing, human oversight requirements) has been provisionally pushed to a later date
- —Multiple US states have introduced or passed legislation specifically targeting automated hiring decisions, though enforcement and scope vary significantly by state
The regulatory direction is consistent even where the details differ: transparency and auditability of automated hiring tools are becoming a compliance requirement, not just a best practice, in the markets this matters most for.
What this does and doesn't mean for your resume
It's reasonable to be concerned about bias in automated screening — the research supports that concern being real, not paranoid. It's less useful to treat every rejection as evidence of algorithmic bias specifically, since parsing failures, keyword mismatches, and genuine qualification gaps remain far more common causes of a resume not advancing than name-based bias in any single application.
What's actually within your control: making sure the resume itself gives an automated system nothing structural to fail on — clean parsing, standard section headers, and field-relevant keyword coverage — so that if a rejection happens, it's not because the system never correctly read your resume in the first place. That's a fixable, checkable problem, separate from the bias question and worth ruling out first.
Check the fixable part first
Career Copilot's free ATS resume checker verifies that your resume actually parses correctly and covers the keywords a screening system would look for — ruling out the most common and most fixable cause of an automated rejection before assuming the cause was bias.