What a 60%+ AI Auto-Reject Rate Actually Means

A 60%+ AI auto-reject rate sounds harsh until you see what’s actually happening behind it. In a typical high-volume campus drive, more than half the candidates who apply never reach a human recruiter — an AI voice screen decides first. That number makes HR leaders nervous the first time they hear it, and it should, because “60% got rejected by an algorithm” is a sentence that deserves scrutiny, not blind trust.

What “auto-reject” doesn’t mean

It doesn’t mean a resume got scanned for keywords and binned. It means a candidate had an actual voice conversation — typically 15 to 20 minutes — and an AI system listened to how they answered, not just what keywords appeared in their CV. That distinction matters because most of what people call “AI screening” today is still resume parsing with an AI label on it. A structured voice conversation is a fundamentally different filter: it tests whether someone can hold a conversation about the role they’re applying for, which a resume can’t tell you.

The four buckets every candidate lands in

Every completed screen gets classified into one of four categories: Strong Go, Go, On Hold, or No Go. “Auto-reject” collapses to the No Go bucket — candidates whose conversation made clear, on communication ability and basic role fit, that they weren’t a match. It is not a single opaque score. It’s a classification with a reason attached, which is the entire point: a rejection that can’t be explained in a sentence isn’t a rejection you should trust, whether it comes from a person or an algorithm.

We built this classification around a specific bet: prioritize communication first, not resume-keyword density. Most rejected profiles aren’t rejected for lacking a credential — they’re rejected because a 15-minute conversation showed they couldn’t yet hold the kind of exchange the role requires. That’s a different failure mode than “wrong college” or “wrong CGPA,” and it’s a fairer one, because it evaluates the actual skill the job needs.

Why 60%+ is the expected shape of the funnel, not a failure of it

Across our pilot work, including a single-evening drive that screened 3,000 candidates for one client in under two hours, and a cumulative 1,099 completed interviews across pilot campaigns, the auto-reject rate has consistently landed at 60% or higher. That’s not a bug in the model. It’s the actual shape of high-volume Indian campus hiring: a role that draws thousands of applicants was never going to have a 50/50 split between qualified and unqualified candidates. When we ran the numbers on screening 3,000 candidates in one evening, the rejection rate wasn’t the surprising part — the fact that it happened in two hours instead of eighteen was.

A screening system that auto-rejects close to zero candidates at that volume isn’t more careful. It’s not actually screening — it’s just deferring every decision to an already-overloaded human recruiter, which is the exact bottleneck AI screening exists to remove.

Think about what a 0% auto-reject rate would actually imply for a role that drew 3,000 applications: every single applicant would need to be equally viable, which almost never happens once a job posting goes out wide enough to reach a big campus pool. The rate isn’t the system being harsh. It’s the system reflecting a funnel that was already lopsided before the AI ever got involved — the same lopsided funnel that used to sit in a recruiter’s inbox for three weeks instead of two hours.

What HR should still check by hand

None of this means the number should go unaudited. We tell every pilot customer the same thing: pull a sample of the No Go bucket every drive and spot-check it. AI voice screening handles accents, code-switching, and nervous candidates better than most people expect, but it isn’t perfect, and pretending otherwise is exactly the kind of overclaiming that erodes trust in the category. Industry research consistently shows most HR leaders now see AI screening as a compliance risk, not just an efficiency play — a concern that only eases if screening vendors get serious about explainability, not defensive about it.

The take

A 60%+ auto-reject rate isn’t the number to be afraid of. The number to be afraid of is a rejection rate you can’t explain to the candidate, your legal team, or yourself. Build the screening funnel so every “No Go” comes with a reason, and the rate itself stops being scary — it just becomes math.

Want to see what your own funnel’s auto-reject reasoning actually looks like? Try the calculator or get in touch.