Candidate Experience in AI Interviews: Why 38% Have Walked Away, and What Keeps Them

Candidates stay in an AI interview when they are told about it up front, it is short, they can take it when it suits them, and a person reviews it and gets back to them. When those things are missing, they leave. In Greenhouse's May 2026 survey of 1,200 US job seekers, 63% had been interviewed by an AI, 38% had walked away from a hiring process because it included an AI interview, 70% were never clearly told up front that AI would be evaluating them, and 46% wanted the option to request a human interview instead.

That is a vendor survey (Greenhouse sells hiring software), and one of the most recent large reads on how candidates feel. The encouraging part is in the same data: only 19% want less AI in hiring. Candidates aren't refusing AI interviews. They are refusing badly run ones. This guide covers what they dislike, and eight design choices that keep them in your process.

What do candidates dislike about AI interviews?

In the same survey, the biggest triggers for walking away were pre-recorded video interviews scored by AI with no human present (33%), companies failing to disclose how AI would be used (27%), and AI monitoring during the process (26%). Each is a design choice, not a property of AI.

Four complaints come up again and again:

• Finding out late. One in five candidates in the survey (21%) only found out that AI would be evaluating them once the interview had started. A surprise at the start of an interview reads as a trick.
• Talking to a camera with nobody there. Pre-recorded video scored by AI, with no human present, was the most-cited reason for walking away. It puts a candidate on camera, with a machine judging and nobody to talk to.
• The machine deciding. In a Pew Research Center survey (fieldwork December 2022), 71% of Americans opposed AI making the final hiring decision.
• Silence afterward. Among Greenhouse's respondents who completed an AI interview, 28% moved forward to the next round, 13% were formally rejected, and 51% never heard back.

That last number deserves more attention than it gets. More than half of the candidates who did what was asked of them never heard back. No interview technology fixes that on its own.

What do candidates want instead?

Guardrails. In the same survey, candidates asked for upfront disclosure (44%), a clear explanation of what the AI is measuring (39%), and the option to request a human interview instead (46%). And 38% wanted to know that a human reviews the AI's evaluation before any decision is made.

When employers get it right, it shows: 38% of candidates came away with a more positive view of the employer, against 34% with a more negative one. An AI interview isn't neutral for your employer brand. It moves most candidates one way or the other.

Eight design choices that keep candidates

Each choice below says why it matters, what to do, and what DIO, the AI interviewer Dio makes, does today. For what candidates see in a DIO interview, see what to expect in an AI interview with DIO.

1. Tell them up front, in plain words

Disclosure belongs in three places: the job posting, the invitation, and the interview page itself. Say that an AI asks the questions, that the interview is recorded, what it covers, how long it takes, who reviews it, and whom to contact. A short AI disclosure statement for an invitation might read:

The first interview for this role is a voice interview with an AI interviewer. It takes about 15 minutes, and you can take it whenever suits you this week. The interview is recorded and transcribed. A recruiter on our team reviews every interview and makes the decision about next steps. If you have questions, or need a different format, reply to this email.

For a job posting, one line will do: "The first interview for this role is a short AI voice interview, reviewed by a person on our team."

The last sentence of the invitation matters more than it looks. With 46% of candidates wanting the option of a human interview, name the person who can arrange one, or another format, for anyone who needs it.

In DIO: the interview page describes it as an AI-guided interview, and before joining, every candidate agrees to a notice that says an AI runs the interview and a person makes the decision. Dio does not send invitations itself; the recruiter shares the link, so the invitation, and what it says, is yours.

2. Keep it short

A first-round interview has a narrow job: confirm the experience on the résumé is real and relevant, find out why the person is looking, and settle the basics. Every question past that adds minutes the candidate has to find. Cut questions the résumé already answers, and save deep technical questions for a later round.

In DIO: most interviews take about 15 minutes.

3. Let them take it any time

Many candidates already have a job, and the best time to interview is often an evening or a weekend. An interview with nothing to schedule removes a round of back-and-forth and the no-shows that come with it. Give a window in your invitation, such as "this week", so candidates know what is expected.

In DIO: there is nothing to schedule. Candidates open the link and take the interview at a time that suits them.

4. Let them repeat or skip a question, and stop

Nerves, a noisy room or a dropped word can wreck an answer. Letting a candidate ask for the question again costs nothing. Letting them skip one is better than forcing a panicked answer, and letting them stop at any time is basic respect.

In DIO: candidates can ask DIO to repeat or skip a question, and can end the interview at any time.

5. Offer captions, and don't require a camera

Captions help candidates who are hard of hearing, who are listening in a second language, or who are in a noisy room. And a camera adds pressure without adding much evidence for most first-round questions; the most-cited walk-away trigger in the survey was pre-recorded video scored by AI.

In DIO: captions are available. A microphone is needed, and a camera is never required.

6. Answer their questions about the role

A first call is two-way. Candidates use it to decide whether they want the job: the pay range, the schedule, whether it's remote, what the team is like, what happens next. An AI interview that only asks questions leaves them guessing, and the good candidates have other options.

In DIO: at the end of the interview, DIO invites the candidate's own questions and answers only from the role's Q&A, which Dio drafts from the job description and the recruiter reviews. When the Q&A doesn't cover a question, DIO says it doesn't have that information and tells the candidate to check with the recruiter. Before you publish a role, make sure its Q&A covers what candidates ask you on first calls.

7. Make sure a person decides, and say so

Candidates ask for this directly: 38% in the survey wanted to know that a human reviews the AI's evaluation before any decision is made. It should be true, not just said. A person should read the evidence, listen to the recording when a score looks wrong, and own the decision. Hard requirements deserve the same care: a knockout question works better as a flag a recruiter reviews than as an automatic rejection.

In DIO: a person on the team records every decision. Knockout questions flag an answer for the recruiter's review; they never end the interview and never reject anyone. Candidates don't see a score.

8. Follow up quickly

More than half the candidates in the survey (51%) never heard back after an AI interview. Fixing that is a team habit, not a feature. Set two standards and hold to them: every interview is reviewed within two business days, and every candidate hears the outcome within a week, including the ones you are not moving forward.

In DIO: the team gets a scorecard with quoted evidence, the recording, the transcript and the candidate's résumé for each interview, plus a PDF summary. The decision, and the message to the candidate, belong to the team.

The eight choices at a glance

What an AI interview can't fix

Some candidate-experience problems sit outside the interview. If the job posting is vague, the pay range is hidden, or three more rounds follow with weeks between them, a well-designed first round won't save the process. And some candidates will always prefer a person. That is a reasonable preference, and the answer is to offer one where you can, not to argue.

There are limits on the tool's side, too, and candidates should hear about them before they start. A DIO interview needs a microphone and is in English only today. Each link works once, one attempt per role, so tell candidates to start when they are ready, and whom to contact if the connection drops.

How to measure candidate experience in AI interviews

You can't improve what you don't count, and none of these numbers needs a survey vendor:

• Completion rate. Completed interviews divided by candidates invited, by role and by where the candidate came from. A drop after a change to the invitation or the plan tells you something.
• Where candidates stop. If many candidates end the interview at the same question, read that question aloud. It is probably confusing, too long or too personal for a first round.
• Time to decision. The days from interview to the candidate hearing the outcome, and the share who heard anything at all. Compare it with the 51% in the survey who never heard back.
• What candidates ask. Read the questions candidates ask at the end of interviews. If many ask about pay or remote work, put the answer in the job posting.
• Two short questions afterward. "Did you know what to expect before you started?" and "Was anything confusing or unfair?", with space to explain. Ask them after the decision is sent, so nobody feels the answer counts. Our candidate experience survey has 25 more, stage by stage.

A checklist before you send the first link

• The job posting says the first round is an AI interview and that a person reviews it.
• The invitation covers how long it takes, that it is recorded, who reviews it, and whom to contact for questions or another format.
• The invitation links to a page that explains what to expect, such as our candidate guide.
• Someone has read the role's interview plan and its Q&A, and the Q&A answers what candidates ask on a first call.
• A recruiter has taken a test interview, start to finish, before any candidate does.
• You have named a person who arranges another format when a candidate needs one.
• You have set the dates by which interviews are reviewed and every candidate hears back.

Candidates don't need an AI interview to feel like a person. They need to know it's an AI, to be treated with the courtesy a good recruiter would show, and to hear back. For how the interview itself works from the hiring team's side, see what an AI interviewer does, and for where each format does better, AI interview vs human interview.