Asynchronous interviews: what they are and what they miss

On 28 September 2026, our AI interviewer took its first live call with a new rule about when an interview is allowed to end. The candidate reached the end of the questions. The interviewer invited their questions, answered the one they asked, said goodbye and tried to close the call. The new rule refused, because the candidate’s last turn had been a question rather than “no more questions”. So the interview carried on, and the interviewer said “Take your time” twice into an empty room until the candidate hung up.

We fixed the rule the same day. But the moment stayed with us, because it showed the one thing every asynchronous interview has in common. Nobody from the hiring team is there. When something goes wrong, there is no recruiter to notice, apologise or end the call politely. Whatever the format does, the candidate experiences alone. That is the whole design problem, and the reason some asynchronous formats are finished by most candidates and others are abandoned.

What is an asynchronous interview?

An asynchronous interview is a first-round interview the candidate completes at a time they choose, with no one from the hiring team present. The candidate answers set questions by recording video, recording audio or speaking with an AI interviewer, and the hiring team reviews the recording, transcript or scores later, on their own schedule.

The word describes the hiring team’s side. For the employer, the interview happens without anyone’s calendar being blocked. For the candidate, the experience depends entirely on the format: speaking to a webcam with a countdown timer is a very different thing from having a spoken conversation that responds to what they say.

Three formats that all get called asynchronous

The formats share a name but not a candidate experience. A one-way video asks a candidate to perform without any feedback; an AI conversation asks them to talk, and responds. We compared the first and the last in more depth in AI voice screening vs one-way video interviews.

Why teams switch to asynchronous interviews

Three reasons come up in almost every conversation we have with a TA team:

• Calendars. A recruiter running live phone screens can only fit so many calls around everything else in the day. An asynchronous first round takes no recruiter time until review, so the number of candidates who get an interview stops depending on how many hours the team has.
• Candidates’ time. Many applicants already work, study or commute during office hours. In our own data, 39.4 percent of 1,705 self-booked interviews were taken outside 9-to-6, as we reported in our hiring data story.
• Consistency. Every candidate gets the same questions in the same order, which is the foundation of a structured interview and much harder to keep up across fifty live calls a week.

The third reason is the one teams underrate. An asynchronous interview is structured by default, because the questions are written once. That is a real advantage for fairness and for comparing candidates, and it is lost the moment the hiring team stops reviewing the results carefully.

There is a cost on the other side, and it is worth naming. In a live screen, the candidate is also interviewing you: they ask about the team, the shift, the manager, and a good recruiter sells the role while they listen. An asynchronous first round gives up most of that. For a high-volume role, where the candidate’s main question is “will I get a fair hearing, and quickly?”, the trade is usually worth it. For a scarce, senior candidate, it often is not.

Why candidates abandon some asynchronous interviews

Candidates do not object to asynchronous interviews as a category. They object to specific experiences. Greenhouse’s 2026 candidate AI interview report, a survey of 2,950 job seekers in five countries, found that the most cited reason US candidates gave for walking away from a hiring process was a pre-recorded video interview scored by AI with no human present, at 33 percent. Close behind was a company failing to disclose how AI would be used, at 27 percent.

Read together, those two numbers describe the void. The candidate records into a camera, does not know who or what will watch, and has no sign anyone is on the other end. Each of those is a design choice, and each can be changed:

• No sign anyone is there. A conversation that responds, even an AI one, is a different experience from a countdown timer.
• No disclosure. Tell candidates before they start that an AI will ask the questions, that a person will review the result, and what is recorded.
• No idea how long. State the length and the number of questions in the invitation.

We saw the effect of removing the video step at one company directly. In one HireQwik pilot, the candidate pool doubled without a video step: 3,000 applications and about 1,700 completed applications with video, then 6,000 and about 4,200 voice-only, at the same company. Those are completed applications, not interview completion rates, and one company is not a study. But the direction matched what candidates tell surveys.

Eight rules for an asynchronous interview candidates finish

1. Keep it to 10-20 minutes. Three to six questions that each need a real answer. Anything that can be a form field (notice period, location, expected salary) should be a form field or a quick yes-or-no question.
2. State the length, the format and who reviews it in the invitation, in the first lines, not the footer.
3. Disclose the AI. If an AI asks the questions or scores the answers, say so plainly before the interview starts. Candidates who find out afterwards feel misled, and a growing number of AI hiring laws now expect it.
4. Let candidates choose a real time. A booking link with slots across the evening and weekend respects the people who cannot interview at 11 AM on a Tuesday.
5. Remind them, then stop. A reminder before the slot and one after a missed slot recover a lot of candidates. In our data, 21 percent of booked interviews never start, which is the number reminders exist to bring down. A long chain of reminders annoys more people than it recovers.
6. Give a way out when something breaks. A candidate whose interview cut out needs a clear message that it did not count against them and a way to book again.
7. Have a person review every decision. Asynchronous should mean the recruiter reviews later, not that nobody reviews. Read the evidence behind each recommendation before rejecting anyone.
8. Offer an alternative on request. Some candidates cannot do a spoken or recorded interview because of a disability, connectivity or circumstance. A short route to a live call costs little and matters a great deal to them.

Reviewing asynchronous interviews without losing the time you saved

An asynchronous first round moves recruiter time from the interview to the review. Done badly, the review eats all of it back: a recruiter who watches every recording end to end has simply rescheduled the phone screens to the evening. The review needs its own routine.

Review in one sitting a day, not as results arrive. Interviews land at 9 PM and on Sunday mornings. Checking each one as it comes in turns the recruiter’s whole day into interruptions. One block each morning, the previous day’s results, is enough for most volumes.

Sort before you open anything. Start with the candidates the evidence says are strongest, because they are the ones a competitor will call first. Then the middle, where judgment matters most. The weakest last.

Read the evidence before the recording. A transcript or a per-question score with a quoted answer takes a minute to read. A recording takes the full length of the interview. Open the recording only where the written evidence leaves a real question: a deal-breaker answer that reads ambiguously, a score you disagree with, a flag.

Spot-check the rejections. The candidates most at risk from an asynchronous process are the ones nobody looks at. Before rejecting a batch, open a handful at random and check that the evidence supports the decision. If it does not, the problem is in the questions or the scoring guide, and it affects everyone in the batch.

Budget the review. For a first round, a few minutes per candidate in the middle band and well under a minute for clear results at either end is realistic once the evidence is good. If the review takes longer than the old phone screen would have, the format is not saving anything, and the place to fix it is what the review screen shows, not how fast the recruiter reads. We showed what that screen should answer first in what HR sees after an AI screen.

Asynchronous or live: when each one fits

Most high-volume teams end up with both: an asynchronous first round for everyone who clears the résumé screen, then live rounds for the shortlist. If you want to see where a first round sits in that sequence, AI interviews for high-volume hiring walks through it.

What the empty room taught us about AI voice interviews

The 28 September call is a good example of what asynchronous design has to get right, because the candidate experienced every flaw in it alone. An AI voice interview is asynchronous for the hiring team and live for the candidate, so the interviewer has to do the things a human would do without thinking: notice when the conversation is over, close it politely, and never end it in the middle of a question.

Here is how a HireQwik interview handles those things today:

• The candidate picks the time. They book their own interview slot from an email link and get reminders before the deadline. Invites and reminders go by email; WhatsApp reminders can be switched on for a job.
• The interview is a conversation. A 10-20 minute AI voice interview in the browser, with nothing to install, with questions from the job and from the candidate’s own résumé, and follow-ups based on what they say.
• Disclosure comes first. Before starting, candidates are told the interview is recorded with video, that the camera must stay on, and that switching tabs or turning the camera off will trigger warnings.
• It ends cleanly. On conversational jobs, since that day’s fixes, the interview closes once the candidate’s questions are answered, or when time is up, and never in the same breath as a new question.
• A person reviews the result. Every interview is recorded and transcribed, and the recommendation (Strong Go, Go, On Hold or No Go) lands within minutes of the call ending, with per-question scores that quote the candidate. The AI recommends; HR decides.

An asynchronous interview is only as good as its worst moment, because that moment happens with no one there to fix it. If you want to judge ours from the candidate’s chair, take the live demo interview. It takes about five minutes, and you get your own scorecard by email.