When Hilton Hotels deployed AI-assisted video interviews for high-volume hourly hiring, they documented something counterintuitive: candidate completion rates rose, and post-hire satisfaction with the process was comparable to traditional in-person screening. A 2023 study in the Journal of Applied Psychology by Langer and colleagues found that candidates who understood how an AI interview worked — and believed it was applied consistently — rated it as more fair than a traditional human interview where they perceived the interviewer might have liked them less for reasons they couldn't control.

The popular narrative about AI interviews is that candidates hate them. The research is more nuanced than that.

What Candidates Fear Before an AI Interview

Candidate anxiety about AI-assisted screening is real, documented, and specific. Research by Van Esch and colleagues (2019) found that candidates' primary concerns fall into three categories:

1. Loss of personal connection. Candidates worry that no human is listening, that their nuance will be missed, and that the interaction will be transactional in a way that disadvantages people who communicate differently from a norm they can't see.

2. Algorithmic opacity. Candidates who do not understand how the AI scores their responses feel powerless. If a score is "unexplained," the concern is that arbitrary or biased factors are being weighted — and that there is no way to know.

3. Privacy and data use. Particularly for voice and video recording, candidates are concerned about where data goes, whether it is being used to train models without consent, and whether their responses will be reviewed by humans they have not met.

These are legitimate concerns. They are also, importantly, concerns that good system design addresses directly — not by eliminating AI from the process, but by being transparent about it.

What Candidates Report After an AI Interview

The post-interview research tells a more positive story than pre-interview surveys would predict.

Anxiety is highest before the interview starts, not during it. A study by Langer, König, and Papathanasiou (2019) found that candidates who completed a structured AI interview reported significantly lower anxiety during the interview than they anticipated beforehand. The structured, predictable format — knowing exactly what will be asked and how the process works — reduces the uncertainty that drives pre-interview anxiety.

Consistency is perceived as fairness. Research on procedural justice in hiring consistently shows that candidates rate processes as fair when they are consistent, relevant to the role, and applied equally to all candidates. Structured AI interviews satisfy all three criteria in a way that unstructured human interviews often do not. A candidate who completed a structured AI screen knows that every other applicant answered the same questions — they were not at a disadvantage because their interviewer was in a bad mood or preferred a different communication style.

67% of candidates prefer structured interviews over unstructured ones when they understand the rationale, according to a meta-analysis of procedural justice research in employment contexts by Hausknecht, Day, and Thomas (2004). Structured AI interviews are, by definition, structured. The format is the function.

Candidate NPS improves when AI interviews are transparent. Internal data from enterprise HR platforms that have deployed AI-assisted screening at scale consistently show that candidate Net Promoter Score — whether they would recommend the employer's hiring process to a friend — improves when the AI interview:

Explains upfront that it is AI-assisted
Tells the candidate what competencies are being assessed
Confirms that a human will review the output before any decision
Provides a clear timeline for next steps

Transparency is not a nice-to-have. It is the primary driver of post-interview satisfaction.

How AI Interviews Can Feel More Fair Than Human Ones

This is the finding that most hiring teams find counterintuitive, and the research behind it is worth understanding.

Human interviewers introduce variability that candidates experience as unfair even when they cannot name it. An interviewer who was distracted by another meeting, who asked different questions to a previous candidate, who responded warmly to one candidate's communication style and neutrally to another's — these variations are invisible to candidates but materially affect outcomes. Candidates intuit this, even without data.

A 2021 study by Gonzalez, Capman, Oswald, Theys, and Tomczak in the Journal of Applied Psychology found that when candidates perceived an interviewer as potentially biased — based on implicit cues including gender, age, or expressed enthusiasm — they rated the interview as less fair regardless of whether bias actually affected the outcome. The perception of a non-neutral interviewer reduces procedural justice scores.

AI-assisted interviews do not carry these interpersonal bias signals. A candidate who might worry that a human interviewer noticed their age, accent, gender presentation, or socioeconomic background cannot attribute those concerns to a structured AI evaluation. The evaluation is what it is: the same questions, the same rubric, the same process as everyone else.

This matters most for groups who have historically experienced discrimination in hiring. Research on diversity and inclusion in AI hiring (Raghavan, Barocas, Kleinberg, and Levy, 2020; published in ACM FAccT) notes that AI systems introduce their own bias risks through training data and rubric design — but also that transparent, auditable, consistently applied AI screening can be more equitable in practice than the unaudited, inconsistently applied human screening it replaces.

The Conditions Under Which Candidate Experience Deteriorates

The research also documents when AI interviews produce negative candidate experiences. Understanding the failure modes is as important as understanding the success cases.

When the purpose is not explained. Candidates who begin an AI interview without knowing it is AI-conducted, or without knowing that a human will review the output, report significantly lower trust and fairness perceptions. Transparency is not optional.

When the format feels disconnected from the role. Candidates who cannot see the connection between the interview questions and the job they applied for rate the interview as less relevant and less fair. Well-designed competency frameworks, with visible alignment to the job description, mitigate this.

When there is no communication after. The biggest driver of negative candidate NPS in AI-assisted processes is not the interview itself. It is the absence of timely, respectful follow-up after the interview. Candidates who complete an AI interview and receive no communication for two weeks experience the process as worse than a traditional phone screen that at least gave them a human interaction. The interview format is not the problem; the follow-up failure is.

When candidates suspect pure automation. If candidates believe a rejection was issued automatically based on an AI score with no human involvement, fairness perceptions drop sharply. This is both an ethical and a legal concern in most jurisdictions. The human-in-the-loop is not just compliance — it is what maintains candidate trust.

What the Research Recommends for Design

The I/O psychology research on candidate experience in AI-assisted hiring converges on five design principles:

1. Disclose before the first question. Tell candidates it is an AI-assisted interview, name the AI component, and confirm human review will occur. Do this before they start, not in the terms and conditions they didn't read.

2. Explain what is being evaluated. Candidates who know the competencies being assessed — and why those competencies matter for the role — feel more prepared and more fairly treated. Opacity breeds suspicion.

3. Keep it appropriately short. Research on candidate experience in structured interviews consistently shows satisfaction declining for sessions over 20–25 minutes. Longer is not more thorough; it is more draining.

4. Follow up promptly and specifically. Even a decline communicated within 5 business days, with a specific reason, produces better candidate NPS than silence followed by a generic rejection. Candidates who feel respected at rejection refer others.

5. Make it mobile-accessible. In India, GCC, and most developing markets, candidates primarily access hiring processes via mobile. An interview format that requires a desktop, a camera, or a fast fixed-line connection systematically excludes a segment of the candidate pool.

Key Takeaways

Pre-interview anxiety about AI interviews is higher than post-interview satisfaction would predict. Candidates fear AI screens more than they dislike them — the actual experience, when well-designed, is often better than anticipated.
67% of candidates prefer structured interview formats when they understand the rationale. AI-assisted interviews are structured by definition.
Consistent, transparent AI interviews can feel more fair than variable human interviews — particularly for candidates who have experienced or perceived bias in traditional processes.
The primary drivers of poor candidate experience in AI-assisted hiring are not the AI itself — they are opacity about the process, disconnection from role requirements, and lack of follow-up communication.
Transparency is the single highest-leverage design variable. Candidates who know what is happening, why, and what comes next report materially higher satisfaction and recommend the employer's process to others.

If you're building or auditing an AI-assisted hiring process and want to see how a well-designed candidate experience is structured from the first screen to the decision communication, the Voxxhire demo walks through each step.

Research citations: Langer, König & Papathanasiou (2019), Journal of Applied Psychology; Hausknecht, Day & Thomas (2004); Van Esch, Black & Ferolie (2019); Gonzalez et al. (2021), Journal of Applied Psychology; Raghavan, Barocas, Kleinberg & Levy (2020), ACM FAccT.

What Candidates Actually Think About AI Interviews (The Research Is Surprising)