Researchers at the University of Washington ran a simple test. They took the same resumes and changed only the names. Resumes with white-associated names were chosen for the next step 85% of the time, while Black-associated names were chosen 9% of the time. Same resumes, different names. Fair enough, you might think. A person will catch that. A newer study tested exactly that. It put 528 people through 1,526 resume screening scenarios. When they had no AI, or an AI built to be neutral, they picked white and non-white candidates at about equal rates. When they worked with a moderately biased AI, their picks tended to follow its lean. Coverage of the study put the agreement at about 90% when the AI was biased. In plain terms, the AI’s bias became theirs. Recruiters Follow AI's Biased Hiring Recommendations 90% of the Time, Research Says This is automation bias, and it works through anchoring. Once a confident list shows up, your brain stops building its own answer and starts judging the AI’s. You feel like you’re reviewing, but you’re mostly agreeing. Here’s a workflow that protects you. I call it “you first, AI second.” Write down your own scoring criteria for the role before you read a single resume. Five criteria is plenty. Score each candidate yourself against those criteria, with names and photos hidden if you can. Only then ask AI for a summary. Make the AI show its evidence. Paste this: “For each candidate, list how they match each of these criteria: [your criteria]. Quote the exact resume line that supports each match. If there’s no evidence, write ‘no evidence.’ Do not rank the candidates. Do not comment on name, age, gender, schools, or employment gaps.” Compare your scores to the AI’s. Where you disagree is where to spend your time. Also note that the researchers pointed out that removing names alone may not solve the problem, so keep checking results over time, not just inputs. AI is useful for reading 200 resumes. It shouldn’t be the first voice in your head.

Derek Mobley says he applied for more than 100 jobs at companies that all used the same hiring platform. His claim is that the rejections came back fast and at odd hours, including the middle of the night, and he points to that speed as evidence that software, not a person, made the call. The vendor, Workday, says its technology doesn’t make hiring decisions. That’s still unproven either way. But in June 2026 a federal judge again refused to dismiss most of the discrimination claims, so the case keeps moving. Now a different story, with a different cause. iTutorGroup’s application software was set up to automatically reject women aged 55 or older and men aged 60 or older, and more than 200 qualified US applicants were turned away. How did anyone find out? One applicant was rejected right away with her real birth date, then reapplied a day later with everything the same except a more recent birth date, and got an interview offer. The company paid $365,000 to settle. (The company denied the allegations in the settlement.) In one case the rule was allegedly written on purpose. In the other, the claim is that the system picked up patterns on its own. Either way, someone got a rejection that no person reviewed, and the employer owned the outcome. Here are the questions to ask your hiring vendor this week: Does any step reject or rank candidates without a human looking? What data was it trained or tuned on? Has anyone tested it for bias, and can you see the results? Which inputs does it use? Ask specifically about age, graduation dates, employment gaps and names. Can you turn off automatic rejection? Who pays if it goes wrong, you or them? Then try the birthday test. Submit two test applications that are identical except for one detail, like a graduation year or a name, and see whether the outcome changes. It’s the same trick that exposed the iTutorGroup software. On the legal side, the rules keep changing, so check with counsel. As of now, New York City’s bias audit rule has applied since July 2023, California’s automated decision regulations since October 2025, and Illinois’ amended Human Rights Act since January 2026. Illinois requires you to notify applicants and employees when AI is used in hiring, promotion, discharge or discipline.