Do hiring managers and job seekers agree on AI fairness?

Greenhouse, a hiring software company, reports a survey in which the two sides of hiring disagree sharply about AI. The headline pairs two figures: 70% of hiring managers say AI helps them make faster and better hiring decisions with fewer recruiter resources, and only 8% of candidates believe AI makes hiring more fair. The headline calls the first figure trust, but the body measures something narrower: a self-report that AI helps. Recruiters sit between the two groups. The excerpt says 50% believe AI has improved hiring overall, "mainly by saving time on screening and scheduling," yet 25% are not confident in their AI systems at all, 8% have no idea what their algorithms prioritize, and only 21% are very confident their systems are not rejecting qualified candidates.

The gap is also a gap in trust over time. The excerpt reports that 46% of U.S. job seekers say their trust in hiring has decreased over the past year, 42% of them blaming AI directly, and that 62% of U.S. Gen-Z entry-level workers have lost trust. On bias, 35% of job seekers think AI has shifted bias from humans to algorithms, and 18% think it has amplified bias by learning from historical patterns. The excerpt's account of the mechanism is brief: candidates face "opaque AI screening they believe they can't beat," while recruiters are "drowning in so many applications." Greenhouse's proposed remedy is not better AI but "re-establishing transparency, introducing 'good friction' like identity verification, and improving the quality of hiring signals." The excerpt says 87% of job seekers consider employer transparency about AI use important, and that such transparency is "largely missing."

The closest library note is the one on overreliance: "augmentation is not inherently safe because overreliance on AI agents can gradually erode the skills and oversight of workers." Greenhouse's recruiter figures are a measured instance of the oversight half of that claim, since many people running screening systems cannot say what those systems prioritize or whether they reject qualified candidates. The credential question sits alongside it. The Do university AI policies actually protect what credentials mean? note argues that permission categories for AI use do not preserve what a credential certifies. Greenhouse reports the same problem from the hiring side: 74% of hiring managers are more concerned about fake credentials, deepfakes, or misrepresented experience than a year ago, and 39% are adding in-person interviews. The Does co-design participation hide misalignment in preference agents? note has a similar shape at a smaller scale, where a group feels well served while a check finds otherwise. The Greenhouse survey measures opinion about outcomes rather than the outcomes, so that parallel is structural, not evidential.

The excerpt does not state the survey's method. It gives no field dates, no sample size for the hiring managers or recruiters (the 1,200 figure covers U.S. job seekers only), no question wording, and no sampling or weighting detail. The job-seeker and employer figures may not come from the same population. The excerpt also does not measure whether AI decisions are actually faster or better; the 70% is what managers say. Chait writes that "our vision is to build AI for hiring," so the survey comes from a party with a stake in how the question is framed. What the data supports is narrower than the headline: the two sides hold very different beliefs about AI in hiring, and the excerpt describes recruiters as more ambivalent than hiring managers. It does not show which side is right.