70% of Hiring Managers Trust AI to Screen Faster. Only 8% of Candidates Call It Fair.

A Greenhouse survey of 4,136 people across the US, UK, Ireland, and Germany, roughly 2,900 job seekers and 1,236 recruiters and hiring managers, found that 70% of hiring managers trust AI to make faster, better hiring decisions. Among the job seekers in the same survey, only 8% said AI makes hiring more fair (Greenhouse newsroom, November 2025). Neither number is the risk on its own. A hiring team confident in its tools isn’t a problem, and a skeptical candidate pool isn’t unusual for any new hiring technology. The risk is the 62-point gap between them, because that gap is where compliance exposure and candidate-experience damage both live.

Where the gap actually shows up

The same survey found 46% of US job seekers report decreased trust in hiring over the past year, 42% directly blame AI for eroding their confidence, and 35% believe AI has simply moved bias from humans to algorithms rather than removing it. For campus hiring specifically, the most relevant number in the whole survey is this one: 62% of US Gen-Z entry-level workers say they’ve lost trust in the hiring process. That’s the exact demographic a campus drive is screening, and it’s the demographic least likely to have been through enough hiring cycles to have a baseline for what “normal” screening friction looks like.

Nobody on the hiring-manager side of a 70%-trust number is lying about their own experience. AI screening genuinely does move faster and catch things a fatigued phone-screener misses. The gap isn’t a disagreement about whether the tool works. It’s a disagreement about whether the process is legible to the person it’s being used on, and legibility is a completely different problem than accuracy.

The recruiter-side number that matters more than the headline stat

Buried further in the same research: only 21% of recruiters say they’re very confident their own screening systems aren’t rejecting qualified candidates by mistake, and 25% admit they lack confidence in their AI systems entirely. Read that next to the 70% hiring-manager trust figure and a more precise picture appears. Hiring managers trust AI to make hiring faster. A meaningful share of the recruiters actually running the system don’t trust it to make hiring accurate. Those are two different kinds of trust, and conflating them is exactly how a company ends up defending an auto-reject rate in front of a regulator or a candidate’s LinkedIn post without ever having checked what it was actually rejecting. We’ve written about what closing that specific confidence gap looks like in practice: auditing every no-go batch by hand rather than trusting the score alone.

What actually narrows the gap, and what doesn’t

Explaining the technology better doesn’t move the 8% figure much. Candidates aren’t distrustful because they don’t understand how a screening model works; they’re distrustful because they can’t tell what happened to their specific application. The lever that moves candidate trust is disclosure paired with a real path to a human, not a better FAQ page about the AI. Tell a candidate plainly that a voice-AI conversation is part of the process before they take it, not after. Give a rejected candidate a route to a person if they want to contest the outcome, even if most candidates never use it. Colorado’s rewritten AI-hiring law, Illinois’s disclosure statute, and NYC’s Local Law 144 all converge on some version of exactly this, which we’ve written about before — the requirement that survives every rewrite is disclosure plus an accessible human override, and it happens to be the same thing that would close a chunk of this trust gap even without a law requiring it.

The honest limit

Disclosure and audit trails won’t close a 62-point trust gap to zero, and claiming they would be overselling. Some of that gap is candidates reacting to a genuinely opaque hiring market, not to AI screening specifically; some of it predates any AI tool a company has deployed. What disclosure and a real human-override path do is close the part of the gap that’s actually about the AI, which based on the survey’s own breakdown (35% specifically citing algorithmic bias, 42% specifically blaming AI) looks like a meaningful chunk of it, not a rounding error.

If your team is running high-volume screening and hasn’t audited what your own rejected pile actually looks like recently, that’s the fastest way to find out which side of this gap you’re actually on. Talk to us about what a disclosure-first, audited screening flow looks like for your funnel.