Do Candidates Accept AI Interviews? What Employers Should Expect

Most candidates will take an AI screening interview, and a meaningful minority resent it. What decides which group you get is not the technology, it is the design: whether you disclose the AI up front, whether the interview responds to what the candidate says, and whether a human replies afterwards either way. Teams that get this wrong see drop-off and angry reviews. Teams that get it right find candidates prefer it to waiting three weeks for a phone screen that never comes.

This is the objection every hiring team raises before buying, and it deserves a straight answer rather than a vendor's reassurance. Here is what actually happens, where the resentment comes from, and what to change so you do not earn it.

Do candidates accept AI interviews?

In practice, acceptance splits along a line that has very little to do with AI itself. Candidates accept an AI interview when it feels like an opportunity they would not otherwise have had. They reject it when it feels like a wall built to keep them away from a human.

Consider the two situations from the candidate's side. In the first, they applied to a job three weeks ago and heard nothing, because a recruiter with 400 resumes never got to theirs. In the second, they applied yesterday and were invited to a 12 minute interview they could take at 9pm after their shift, where they got to actually explain why the gap in their resume exists. The second is not worse than the first. The first is not a human process, it is an absence of one.

The complaint that shows up in Reddit threads and Glassdoor reviews is almost never "I had to interview with an AI." It is "I recorded answers into a void and never heard back." Those are different failures, and the second one is entirely fixable without removing the AI.

Where the resentment actually comes from

The one-way video void

The format that generated most of this bad feeling is asynchronous video: a question appears on screen, a countdown starts, the candidate records an answer to nobody. There is no acknowledgment, no chance to clarify what the question meant, no way to recover if the first sentence came out wrong, and often no reply for weeks.

Candidates describe it as being processed rather than interviewed, and they are not wrong. The format takes the least pleasant part of interviewing, the feeling of performing for an unseen judge, and removes the one thing that makes it bearable, which is another human reacting. We have written about where this format does and does not fit on our one-way video interview software page, and the honest summary is that it is cheap and it costs you candidates.

The suspicion of being scored on the wrong things

The second source of resentment is a reasonable fear that the machine is judging their face, their accent, or the noise of their apartment rather than their answers. That fear has a real basis. Some early video interviewing products analyzed facial expressions and vocal tone, several of those features were quietly retired after researchers questioned whether they measured anything at all, and candidates have long memories.

If your screening scores the substance of the answer against a written rubric and nothing else, say so explicitly, in the invitation, in plain language. Candidates who know they are being judged on what they said, not how they looked saying it, behave completely differently in the interview.

The dead end

The third one is the simplest and the most self-inflicted. A candidate completes the interview, and then nothing happens. No confirmation, no timeline, no outcome. Whatever goodwill the format had is gone, and the review they leave will mention the AI, even though the AI is not what upset them.

What candidates actually like about it

Once you remove those three failures, the format has genuine advantages that candidates will tell you about unprompted.

• It happens on their schedule. Anyone currently employed has to take a phone screen in a stairwell on their lunch break. An interview they can take at 9pm on Sunday is a real improvement, particularly for shift workers and caregivers.
• Everybody gets one. A candidate whose resume would have lost the keyword lottery gets to make their case out loud. Career changers and people with non-linear histories benefit most, because a resume is the format that serves them worst.
• No small talk lottery. A structured interview asks everyone the same job-relevant questions. Candidates who are not naturally good at charming a stranger in the first 90 seconds are no longer punished for it.
• It is fast. An answer within days rather than a silence lasting weeks. Speed reads as respect, and candidates say so.

Does an AI interview increase candidate drop-off?

It can, and the variables that drive it are within your control. Length is the biggest one: completion falls sharply past the 15 minute mark, particularly for hourly roles where the candidate is applying to several employers from their phone. Keep the screen between 8 and 15 minutes.

The second variable is what you ask people to install or set up. Every additional step, an account to create, an app to download, a browser permission to grant, costs you candidates who were perfectly willing to be interviewed. The third is whether the interview responds. A conversation where the agent asks a follow-up based on what the candidate just said holds attention. A recording light does not.

The comparison to hold in your head is not AI interview versus perfect human interview. It is AI interview versus the automated rejection email those candidates were realistically going to get. Drop-off from an AI screen is worth measuring, but the drop-off from never being screened at all is 100 percent, and most hiring teams have simply stopped counting it.

How to run an AI interview candidates do not resent

Disclose it before they start, not in a footer. Say that the first-round interview is conducted by an AI interviewer, say what it will ask about, say roughly how long it takes, and say that a human reviews the result. Get consent. In New York City this is a legal requirement under Local Law 144, and in Illinois for AI video interviews, but it is worth doing everywhere because it is the difference between transparency and a trick.

Tell them what is being scored. "Your answers are scored against the same job-related criteria as every other candidate, by a rubric we defined in advance" removes most of the anxiety about facial analysis in one sentence.

Make it a conversation. An agent that listens and follows up gives the candidate a chance to recover from a weak first answer, which is exactly the chance a recording denies them. It also gets you better signal, so this is not a concession, it is the point.

Let them ask questions. Candidates want to know about the shift pattern, the pay, whether the role is really hybrid. If those questions are answered during the interview, or by an assistant that can answer them instantly on your careers page, the process stops feeling one-directional.

Reply to everyone. Every single candidate who completed an interview gets an outcome, within days, and the ones who are not advancing are told. If you screen 400 people with AI and reply to 15, you have automated the part that helps you and skipped the part that helps them, and candidates can tell.

Keep a human in the decision. Not just for the law, though EEOC guidance and Local Law 144 both assume a person is accountable. A candidate who knows a recruiter read their transcript accepts the outcome. A candidate who believes a machine rejected them tells people about it.

What about candidates using AI to cheat the interview?

This is the mirror-image worry, and it comes up in the same conversation. Yes, some candidates will have a language model open in another window. In a structured interview conducted by an agent that asks unscripted follow-ups about the specific example a candidate just gave, this tends to collapse quickly: the first answer sounds polished and the second one, about the detail of the outage they claimed to have led, does not.

It is also worth remembering that a first-round screen is not a test with correct answers. It is an assessment of whether someone has done the thing they say they have done. Fluent generic answers score badly against a rubric anchored in specifics, which is one more reason to write the rubric around concrete evidence rather than around how good the answer sounded.

The honest summary

An AI interview is not warmer than a good recruiter. Nothing is, and any vendor claiming otherwise is lying to you. What it is warmer than is the process it actually replaces at volume, which is silence: the resume nobody opened, the rejection email nobody wrote, the candidate who never found out why.

Disclose it, make it conversational, score the substance, and reply to everyone. Do that and candidate reaction stops being your biggest risk in this purchase, and starts being one of the reasons you made it. That is how our AI video interview is designed to run, and the compliance side is set out in our guide to the NYC Local Law 144 bias audit.