Who is concerned about AI-assisted job candidates? Turns out, everyone is.

• 99% of HR teams are concerned about AI-assisted candidate fraud
• 77% are confident they can detect it (only 18% are very confident)
• Yet in the past 12 months: 40% saw concerning AI use in interviews, 37% suspected proxy interviewing, 33% found fraudulent resumes
• 60% of HR leaders report interview no-shows are happening more often than two years ago
• The gap: universal alarm paired with false confidence means fraud is getting through anyway

A confidence paradox

Seventy-seven percent of HR leaders are confident their hiring process can detect AI-assisted candidate fraud.

That's a lot of confidence.

But here's what happened in the past 12 months: forty percent of teams saw candidates use AI in ways that raised concerns during interviews. Thirty-seven percent suspected someone else was actually doing the interview. Thirty-three percent found fraudulent resumes.

So the question isn't why people are overconfident in general. It's why they stayed confident after their process already failed to catch fraud.

Here's what the data shows: ninety-nine percent of teams worry about fraud, seventy-seven percent say they can catch it, and forty percent have already seen it happen anyway.

The confidence didn't drop after they encountered fraud. That's the real puzzle.

Most people would expect that encountering something you're supposed to catch would shake your confidence in your ability to catch it. But these teams reported both the concern and the confidence in the same survey. Their certainty survived their own experience.

That gap—between "I'm confident this works" and "this already didn't work"—is what stops teams from rebuilding. If you believe your process is sound, you add a step. If you think it's fundamentally broken, you change it.

No-shows reveal a broken signal

Sixty percent of HR leaders report that candidates not showing up for scheduled interviews is happening more often now than it did two years ago.

That's not a minor uptick. That's a fundamental shift.

Two years ago, when someone booked an interview, they usually showed up. The friction in the application process meant people who got that far were genuinely interested. Now they're not.

The application process changed. One-click tools let candidates apply without reading the job posting. Auto-apply services send applications while someone's in a meeting. AI polishes resumes during conference calls. The 60% increase in no-shows is the result: people are reaching the interview stage without ever intending to be there.

When your application pool is mostly low-intent submissions, your interview stage catches people who applied on autopilot, not people who actually want the job. Some got other offers and ghosted. Some were never real to begin with. All of them made it through your screening.

That's the real problem. It's not whether your fraud detection catches the fakes. It's that your process can't rebuild signal in an application environment where signal has already collapsed.

How to rebuild signal

The fix requires two things: filtering for real intent much earlier in the process, and handing off the volume screening to someone built to handle it.

1. Filter for intent at the front.

Real interest shows up in behavior. Candidates who reached out to a real person with a thoughtful message. Candidates who applied to one well-matched role, not a scattershot of your openings. Candidates who reference something specific about your team, product, or problem you're solving. Candidates who did unrequested work—a sample project, a teardown, a proposal.

These signals are harder to fake at scale. They require actual attention. They're the opposite of one-click applications.

Knockout questions early in the process can filter these in fast. So can requiring a real touchpoint—a phone screen, a brief video submission—before candidates move forward. These steps thin a flooded pool and screen out spray-and-apply submissions before they consume time.

2. Hand the screening to a recruitment partner.

Ninety-three percent of teams have updated their hiring practices in response to AI concerns. But only 22% describe those changes as significant. Most teams are trying to patch an overloaded process with a few extra steps.

A recruitment partner built for high volume can do the sorting, verifying, and vetting, then deliver HR a short list of real, interested candidates. That frees your team to do what only they can do: have actual conversations with people who might work out.

The burden doesn't belong on HR. The solution isn't better fraud detection. It's a smaller, real, verified pool of genuinely interested people in front of someone who can actually evaluate them.

Download our most recent report, When effortless applications made hiring harder.