Can AI resume screening filter out qualified candidates unfairly

A Well-Documented, Real Concern

Yes, AI resume screening can genuinely and unfairly filter out qualified candidates, and this isn’t merely a theoretical risk — documented cases and research have shown specific, real ways this can happen, making it a well-established concern that has prompted both industry attention and specific regulatory responses.

Overly Rigid Keyword Matching

One documented cause is overly rigid keyword matching, where a system looks for specific exact terms and fails to recognize genuinely equivalent experience described using different, but substantively similar, terminology — a candidate with directly relevant experience described using slightly different language than the system’s defined keywords might score poorly despite being genuinely well-qualified.

Bias Learned From Historical Hiring Data

Systems trained on a company’s historical hiring data can inadvertently learn and replicate past biased hiring patterns present in that data. A widely reported example involved a major technology company discontinuing an internal resume screening tool after discovering it had learned to systematically downgrade resumes containing indicators associated with female candidates, reflecting historical gender imbalances in that company’s prior hiring data rather than any genuine measure of qualification.

Resume Parsing and Formatting Errors

Some qualified candidates get filtered out not due to any judgment about their qualifications at all, but because of technical parsing errors — resumes using complex formatting, tables, graphics, or unusual document structures can sometimes be incorrectly read by automated systems, causing genuinely relevant information to be missed or misinterpreted even though it’s present on the actual document.

Why This Has Prompted Specific Regulatory Responses

Given these documented risks, some jurisdictions have enacted specific regulations requiring bias auditing or additional transparency for automated employment decision tools, reflecting a policy recognition that these systems can produce genuinely unfair outcomes without proper design and ongoing oversight.

Why Responsible Employers Take Steps to Address These Risks

Employers using these tools responsibly generally conduct their own bias testing, maintain some form of human review for automated screening decisions, and periodically audit system outputs for concerning patterns, recognizing that without these safeguards, real, qualified candidates can be unfairly excluded through no fault of their own.

Bottom Line

AI resume screening genuinely can unfairly filter out qualified candidates, through documented mechanisms including overly rigid keyword matching, bias learned from historical hiring data, and resume parsing errors related to formatting — a well-established, real concern that has prompted specific regulatory responses and responsible-use practices rather than a purely hypothetical risk.

Go deeper

• What Real-World Harms Have Resulted From Biased AI Systems?