It Isn't Just You: Candidates Object to Losing Influence

How Common Is the Reaction?

Common enough that several research groups have measured it, in more than one country, and found it pointing mostly the same way. A review of applicants' procedural-fairness perceptions of algorithmic recruitment tools, built on seven studies with more than 1,300 participants between them, reported overall fairness perceptions mixed but perceptions of control and social presence mostly negative ¹.

Participants, mostly in scenario studies rather than real applications, felt less able to influence the outcome and felt the human element missing.

Surveys that put the question directly to people attach numbers to it. In the Early Careers Survey 2026, a majority of respondents said employer use of AI was unfair at every stage of recruitment except initial qualification screening, and discomfort ran highest around online interviews and virtual assessment centres. Read from the report's own chart, 83% called AI use in an online recorded interview unfair ³. Neurodivergent respondents were more uncomfortable than their peers at every stage.

Where AI touches the interview is where the objection concentrates.

That question asked what people think is fair, not what happened to them, and the sample was self-selected UK early-careers respondents. The report's free-text answers record a related complaint: that recorded interviews get reviewed by systems rather than by a person.

Willingness moves the same way. In a Pew Research Center survey of 11,004 US adults, fielded in December 2022, 66% said they would not want to apply for a job with an employer that uses AI to help make hiring decisions, and 32% said they would ⁴. That is a stated intention rather than a measured behaviour: nobody in the sample was standing in front of a job they wanted.

The question also covered AI that helps make hiring decisions, which is broader than AI that makes them.

What Are Candidates Objecting To?

Losing influence over the outcome, and the absence of a person. Those are the two findings that keep recurring, and neither is a claim that the system gets the answer wrong ¹. A 2026 study of asynchronous AI interviewers reached the same place from a different direction, working from 11 subreddit discussion threads and then interviews with 17 applicants ⁵.

What it found was a gap between what applicants expected, shaped by how familiar they were with language models, and how employers framed the exercise. That unmet expectation damaged their sense of agency and trust in the process, and pushed some of them toward workarounds ⁵. Read those workarounds as something an unaccountable one-way format provokes, rather than as a fact about applicants. It is an arXiv preprint, not peer reviewed when it was checked.

The qualitative half rests on 17 interviews and self-selected Reddit posts, and its second phase, with 180 participants, tested a research prototype rather than a product an employer would send you.

Explaining how the system decides does not reliably improve how candidates perceive it ¹.

One reading of all this points at the employer rather than at you. A first round with nobody in it says something about how the process was designed rather than about you, which is the ground whether an AI first round is a bad sign covers. What the research supports is narrower than a verdict on the company: the format costs an employer some goodwill, and the pooled data put most of that cost on one-way and AI-powered formats.

Which Formats Carry the Penalty?

One-way recorded and AI-powered interviews, more than remote interviewing as such. A 2026 open-access meta-analysis pooled 24 studies, 31 independent samples and 3,874 participants comparing technology-mediated interviews with face-to-face. Applicants in the technology-mediated ones reported modestly lower organisational attractiveness and weaker behavioural intentions, while justice perceptions and negative emotional responses did not differ by mode ⁶.

Split by format, the picture sharpens. Asynchronous interviews correlated negatively with justice perceptions and with behavioural intentions, and interviews using AI-powered technology correlated more strongly negatively with organisational attractiveness and with intentions ².

Live technology-mediated interviews did not, in that same pooled data.

So what people react to looks more like the one-way part and the scoring than the screen itself.

Those splits rest on small numbers of studies, several with only three to five behind them, and the authors describe the moderation as partially supported ². The AI-powered grouping also covers whatever the primary studies called AI across roughly six years, which is not one technology.

Your own chances are not what any of this measures. These are correlations with attitudes and intentions rather than with outcomes. The same meta-analysis found interview mode had a near-zero relationship with interview ratings, on seven samples with a wide interval ⁶. Read that as not detectably different, not as proven equal, and read the whole picture as a statement about how a format feels rather than about how it rates you.

Ask Who Reviews It, and Where the Live Conversation Is

Ask where in the process you get to talk to a person, and ask who reviews the recording. Both are ordinary questions, and the answers speak to exactly what the research keeps finding: influence over the outcome, and whether a person is present at all ¹. An employer that has thought about the format will have an answer ready, and there are other questions worth asking about how a company uses AI in hiring.

A request for a different format is a different act from turning the interview down, and it costs nothing to make before you decide anything. Turning down an AI interview covers what usually happens when you decline outright, and why a flat refusal is rarely the strongest version of the move. If the discomfort underneath the question is really about whether a human ever watches an AI-scored interview, that question has its own, more specific answer.

In one small study, people who had actually sat one were less uniformly negative than the scenario studies suggest.

Eighteen Korean applicants in their twenties who had recently completed an AI video interview told researchers they found it better than traditional evaluation procedures on procedural fairness, objectivity and consistency ⁷. Some of the same applicants were still dissatisfied at being assessed by AI at all, and the digital divide came up repeatedly across the study's groups. It is a qualitative study on one platform in one country, so it records what those people said and not how many feel each way.

Disliking the format is not a failure to adapt.

It is not, on its own, a reason to withdraw either. Judge the employer on what it will tell you about the process, and on whether a person turns up in it at some point.

See how it works >