Ask candidates about AI in hiring. Then ask candidates who've used it.

The story everyone tells about candidates and AI deserves a closer look.

‍

Open any HR publication and the consensus is settled. AI adoption is climbing. Candidates resent it. Trust is collapsing. The future of recruitment depends on putting humans back in the loop because applicants will not tolerate being processed by a machine.

The consensus has plenty of evidence to draw on. Gartner's Q3 2025 survey of 2,901 candidates found that 68% prefer human interactions to AI or chatbots, up from 58% two years earlier. Greenhouse's 2025 research across the US, UK, Ireland, Germany and Australia reports that a meaningful share of candidates have withdrawn from a hiring process specifically because it included an AI interview. Fortune ran a piece in August 2025 in which unemployed professionals said they would rather stay out of work than take another bot interview. ICIMS found more than a third of entry-level candidates hesitant to trust AI in hiring, with its impersonal nature the most cited concern.

That looks like a settled question. The picture is more interesting than that.

Two questions, two very different answers

The surveys above ask candidates how they feel about the idea of AI in hiring. A different and growing body of research asks candidates how they felt about an AI experience they actually completed. The two questions produce different answers.

Bullhorn's 2025 Global Recruitment Insights and Data report surveyed nearly 2,800 professionals. 77% of those who had interacted with AI in their job search rated the experience positively. 88% rated AI voice agents as as good as, or better than, a human interview.

A randomised field experiment by Dr. Brian Jabarian at the University of Chicago Booth School of Business and Dr. Luca Henkel at Erasmus University Rotterdam, drawing on more than 70,000 candidate interviews across 48 roles in healthcare, IT and industrial sectors, found that candidate satisfaction was preserved relative to human interviews, that AI-interviewed candidates received 12% more job offers and showed 17% higher 30-day retention, and that gender bias complaints were cut nearly in half. The most telling finding: when candidates were given the choice, 78% chose AI over a human interviewer.

Our own data points the same way. Hundreds of thousands of candidate responses across multiple industries and continents, drawn from voice interviews run by Mochi, our voice screening agent, and skills assessments run by Shiro, our skills screening agent. Not a curated group. Warehouse roles, fashion and beauty retail, hospitality, internships, airport management, executive assistant work, consulting interns. Some had years of interview experience. Many had never used AI for anything before.

The vast majority rated the experience positively. Most said the process felt fair. Most said it improved their perception of the company they were applying to. And in free-text comments left without any prompt to elaborate, large numbers described the experience as conversational, low-pressure, and more human than they expected.

Two questions. Two answers. Worth understanding why.

In their own words

The scores tell one story. The comments tell a sharper one.

"Honestly, it didn't feel like talking to an AI at all. The conversation felt natural, the questions were well-paced, and the overall experience was smooth and professional."

"Even though I was conversing with an AI assistant, it felt like I was talking to a person, without all the added pressure."

"It was more of a conversation than an interview. The atmosphere was relaxed and free of any interview anxiety."

"I loved how realistic the experience was. It made me feel like I was being interviewed by a real person."

"The scenarios tested real sales skills and fashion knowledge in a way that felt fair and relevant to the role."

"It felt like I was being interviewed in a way, and I thought that was quite refreshing compared to other applications I've done in the past."

One theme runs through the qualitative data, and almost nobody predicted it. Candidates told us the AI made them less nervous than a traditional interview, not more. The social performance of being judged by another person dropped away. The interaction became about the substance of what they had to say, not how they were saying it.

The assumption was that candidates would feel surveilled and processed. A meaningful share feel the opposite. They feel they were finally allowed to focus on the work.

The legitimate concerns

The contrarian case is not that AI in hiring is universally loved. Even the strongest data sets contain candidates who would prefer to interact with a person, and that preference deserves to be respected, not designed around.

A few patterns show up consistently in the qualitative data across the industry. Some candidates want a clear path back to a human when they need one. Some take time to adjust to a new interface, especially when it's their first AI interview. Some want to know how they did and what a strong answer would have looked like, feedback that traditional hiring rarely provides either. These are real signals, and they shape how good AI experiences get built.

What the data also shows is that the same candidates who flag these concerns rate the experience positively when the design accounts for them. The preference for human contact doesn't disappear. It stops being a deal-breaker when the experience is conversational, the path to a human is visible, and candidates leave the conversation knowing more about the role than when they arrived.

Why the data disagrees with itself

Two things are going on.

First, AI in hiring is not one thing. Experiences vary widely, and the experiences candidates have had vary even more widely. Some AI experiences are conversational, role-relevant, transparent about what's being measured, and offer a clear path to human contact when something needs it. Others don't. Candidates who go through the first kind don't generalise to "AI in hiring is fine." They generalise to "this AI experience worked for me." Candidates who go through the second kind generalise the opposite way. Both reactions are rational responses to different products. They get reported as if they were responses to the same one.

Second, the candidates being asked in headline surveys are often not the same candidates doing the assessments. A "would you prefer to interact with a human or an AI" question gathers a hypothetical reaction from people, many of whom have never been through a well-designed AI experience. Hypothetical reactions are almost always more negative than actual reactions. The hypothetical is about an abstract loss of dignity. The actual is about whether the interaction made sense and respected your time.

Well-designed AI assessments respect candidates' time. That's a meaningful contrast with a traditional recruitment baseline that has its own challenges. Talent Board's CandE benchmark research, drawing on more than 1.4 million job seekers since 2011, has long documented candidate frustration with the responsiveness and consistency of traditional hiring processes. Worth weighing that baseline before assuming pre-AI hiring was working well for candidates.

It is also worth noting where the most prestigious firms in the world are placing their bets. McKinsey has piloted its own AI interview as part of its US final-round process, with broader rollout expected in 2026. BCG publishes an AI Talent Promise that sets out how the firm uses AI responsibly in its hiring. Deloitte's 2025 Global Human Capital Trends, surveying more than 13,000 leaders across 93 countries, frames AI in talent acquisition as a tension to be designed for, thoughtfully and with intent.

The firms candidates aspire to work for are leaning into AI in hiring, with care.

The signal underneath the noise

The candidates who have actually been through these experiences are telling us something specific and useful. Design matters more than the presence of AI.

The experiences candidates rate highest tend to share a few characteristics. They feel conversational rather than interrogative. The questions are grounded in real scenarios from the actual job. The AI sounds and behaves naturally enough that the interaction is about content, not medium. The process is short. It works at any hour, in any time zone, from a phone. Fairness is visible, not just claimed. And when something needs a human, there's a clear path to one.

These are design choices, and they're within every employer's control.

The argument is not "use AI" or "don't use AI." The argument is: if you are going to use AI, invest in designing the experience well, and the candidates will tell you whether you did.

The honest read

The story that candidates uniformly reject AI in hiring is incomplete. The story that candidates respond to how AI is designed is well supported by the data. So is the story that candidates respond to how any hiring process is designed, with or without AI.

When the design is right, the data is consistent. Hundreds of thousands of candidates across multiple research bodies, vendors, and industries are reporting positive experiences, with satisfaction in the same band as the best consumer brands. Many of them are spontaneously telling us the AI experience felt more human, more relaxed, and more fair than what they had been through before.

The conversation about AI in hiring would be more honest, and more useful, if it started from that fact. Candidates can engage with the technology. They are. The question worth asking is whether the employers using it are giving them an experience that earns it.

‍

It's the question we ask ourselves everyday at Maki.

‍