The gatekeepers aren't human anymore, and when they fail you, there's no one to call.
The Summary
- Nearly 40% of job candidates now abandon hiring processes that require AI interviews, and those who stay often hear nothing back
- One medical student spent six months reverse-engineering the algorithm he believed rejected him, using Python to probe the black box between his resume and silence
- The human cost of automated hiring: you can't argue with a model, you can't ask for feedback, and you can't know if you were ever really considered
The Signal
A medical student with credentials strong enough to land interviews pre-pandemic suddenly couldn't get callbacks. Same resume, same schools, different result. The only variable that changed: companies started outsourcing first-pass screening to AI. So he did what any technically literate person with time and rage would do. He learned Python and started testing the system.
Six months of experimentation. Tweaking keywords, reformatting sections, A/B testing his own life story against an opaque algorithm. Not to game the system, exactly. To understand whether he was being judged at all.
"The injustice isn't just the rejection. It's that you can't know if the rejection was legitimate."
This isn't one person's paranoia. The Fortune data shows 4 in 10 candidates are walking away from jobs that require AI interviews. That's not resistance to technology. That's a signal that the technology broke something fundamental about hiring: the human transaction of being seen and evaluated by another human.
The candidates who do submit to AI screening? Many report complete radio silence afterward. No rejection email. No feedback. Just the algorithmic equivalent of being ghosted. When a human recruiter passes on you, there's at least the possibility of learning why, of improving, of understanding the rules. When an AI passes, you're left guessing which words, which formatting quirks, which invisible pattern triggered the filter.
Here's what makes this worse than old-school resume keyword stuffing:
- AI models are proprietary black boxes. You can't FOIA them or audit them like you could a government hiring rubric.
- The models are often trained on historical hiring data, which means they encode the biases of previous human decisions at scale.
- Companies adopt these tools to "remove human bias," but they've actually just made bias unaccountable and unreviewable.
The medical student's approach is telling. He didn't just complain. He built a testing apparatus. He treated the hiring AI like a scientific problem and tried to measure its behavior empirically. That kind of agency shouldn't be required to apply for a job.
The Implication
If you're hiring, understand that requiring AI-mediated screening is now a candidate experience tax. You're losing 40% of applicants before you even see them. If your talent pool is deep enough that you don't care, fine. But if you're competing for skilled people, this is a unforced error.
If you're applying, document everything. Track where you apply, what format you use, what happens next. If you have technical chops, consider what this medical student did: treat the system as a reversible problem. The models aren't magic. They're just math, and math can be studied. The companies deploying them are betting you won't bother. Prove them wrong.