The story everyone is telling online and offline seems to carry a common thread: candidates hate AI in hiring, full stop. But the data says otherwise. And it’s a lot more interesting than that simple blanket statement.
What the data shows
Recent industry research on candidate sentiment found that only 19 percent of candidates want less AI in hiring. The majority want the same amount or more. What they want alongside it is guardrails: upfront disclosure that AI is being used, at 44 percent; a clear explanation of what the AI is measuring, at 39 percent; confirmation that a person reviews the AI's evaluation before any decision gets made, at 38 percent; and evidence the tool has been checked for bias, at 29 percent.
That pretty solidly disarms the argument that they’re rejecting the technology itself. It’s a more reasonable request they’re making. A request for the same things anyone would want from a decision that affects their livelihood: know when it's happening, understand what it's measuring, and know a person is still involved.
Where the anger shows up is different, and it's specific. A separate industry report found that nearly four in ten candidates have withdrawn from a hiring process specifically because it included an AI interview, with more saying they'd do the same if required to participate again. And given the data, we can clearly see that they’re not leaving because of the involvement of a model, but rather, the lack of disclosure of said model. Because if it wasn’t disclosed, they have no control, no knowledge of the process. What is it looking at? Is there a human involved at any point?
The black box is the real problem
The pattern more worth our time, and a much more nuanced one at that, is: candidates who haven't yet gone through an AI-driven screening process report worrying about it more than candidates who already have. The concern isn't the technology itself. It's the absence of an explanation. A faceless process making a tremendously consequential call, with no visibility into how, is a reasonable thing to distrust, regardless of what's actually running underneath it.
This is worth saying plainly: an assessment tool that never tells anyone how it works, what it's measuring, or who reviews the result before it counts against someone is careless. It’s of course, quite convenient for companies that invested in the tool. But extremely opaque to the person it’s being used on.
What a design built around this should look like
This is something we take very seriously, not only in our day-to-day, but from the moment we started building our product. The system doesn't decide anything. It produces information, a profile, for a person to use in making a decision, and that's true even where the stakes are highest. Where risk is flagged, the result is capped and routed for human review, regardless of how everything else scored. A person is the one deciding, every single time. No exceptions.
The takeaway
Like we said, it’s clear that candidates aren't mad that hiring is now heavily reliant on AI. But the fact that it showed up without explanation and rarely (if ever) asks for their consent shouldn’t surprise anyone that it’s making people unhappy to say the least.
The good news is that it’s an entirely solvable problem. We in the industry are responsible for building tools that make hiring and all the humans involved in it better, not worse.
Take a look at how our process starts with voice-AI science.
