In short
- Phone screen: a live, personal call that can follow up and sell the job, but costs recruiter time, is hard to schedule and varies from call to call.
- One-way video interview: candidates record answers to fixed questions when it suits them. Consistent and easy to schedule, but no follow-ups, and someone still watches every recording.
- AI interview: a two-way conversation with an AI interviewer that asks follow-ups, then a report with evidence. Scales well; some candidates distrust AI, and a person must check and decide.
- Research: the format changes ratings and candidate reactions, so use one format for everyone at a stage. Structure, not the format, is what predicts performance.
- Combine them: one format for the first round, people for the final round and the offer, and a person for anyone who asks.
Choosing a first-round format
AI interview vs one-way video interview vs phone screen: most hiring teams pick one for the first round without comparing them. This guide sets them side by side, with what the research says, and ends with when each fits and how to combine them.
Three ways to run a first-round interview
All three decide, early and at scale, who goes on to a longer interview. They differ in who asks, whether anyone listens live, and how much work lands on the recruiter.
The phone screening interview
A recruiter calls for 15 to 30 minutes, checks the basics (salary, location, notice period) and asks a few questions about the role, live: they can follow up, answer questions and sell the job.
- Good at: a real conversation, answering the candidate's questions, reading the context of an unusual career, and building a relationship with candidates you want to win.
- Weak at: scheduling (every call needs a slot both people can make), consistency (each recruiter asks different things, and the same recruiter drifts over a week), and records (notes are as good as the person taking them). Recruiter time grows with every candidate.
The one-way (asynchronous) video interview
The candidate gets a link, reads or watches each pre-set question, and records an answer on camera, usually with a set preparation time and a limit on retakes. Nobody is on the other side. Reviewers watch the recordings later. That's why it's also called an asynchronous video interview.
- Good at: scheduling (candidates record when it suits them), the same questions in the same order for everyone, and a recording reviewers can share and rewatch.
- Weak at: follow-up questions (there are none: a vague answer stays vague), and reviewer time (someone still has to watch every answer). Candidates often find talking to a camera with no one there awkward, and a camera-only format needs an alternative for people who can't or won't use video.
The AI interview
An AI interviewer holds the conversation: it asks your questions by voice or in text, listens to the answer and asks follow-up questions when an answer is thin, within the limits you set. Afterwards, an analysis compares the answers with the points you said a good answer covers and shows the evidence.
- Good at: scheduling (candidates take it any time before the deadline), consistency (the same questions and the same criteria for everyone), follow-ups on every answer, and reviewer time (you read a structured report with quotes instead of replaying every minute).
- Weak at: the human touch. Some candidates dislike being interviewed or evaluated by AI, the AI can't sell the job or judge an unusual case the way a recruiter can, and transcripts and analyses can be wrong, so a person has to check the evidence. It also brings the most legal duties of the three.
AI interview vs one-way video interview vs phone screen, side by side
| Phone screen | One-way video interview | AI interview | |
|---|---|---|---|
| Scheduling | A slot for both people; rescheduling is common | Any time before the deadline | Any time before the deadline |
| Same questions for everyone | Only if the recruiter sticks to a script | Yes, fixed in advance | Yes, set in advance |
| Follow-up questions | Yes, as the recruiter sees fit | None | Yes, on every answer, within your limits |
| Candidate can ask questions | Yes | No | Usually only about the task |
| Reviewer time | The whole call, live | Watching every recording | Reading a report with quotes; recordings to check |
| Depth | Depends on the recruiter | Shallow: one answer per question | Deeper than one answer, less than a skilled interviewer |
| Consistency of judgment | Low unless scored against set criteria | Depends on whether reviewers use a scorecard | Same criteria for everyone; a person checks |
| Candidate reactions | Familiar | Often more negative than live formats | Mixed; many people distrust AI in hiring |
| Accessibility | Hard for deaf candidates or people with a speech impairment without an alternative | Camera required; needs an alternative | Depends on the tool: look for text mode and extra time |
| Recruiter judgment | Built in | Later, from recordings | Later, from evidence; the team decides |
| Legal notice duties | General data protection | Extra duties where video is analyzed by AI (e.g. Illinois) | AI hiring rules (EU AI Act, NYC, Illinois, Colorado, California) |
No format wins every row: the phone screen wins on relationship, the one-way video on simple scale, the AI interview on scale with follow-ups.
What research says about the formats
- Mode changes ratings. A meta-analysis of 12 studies (Blacksmith, Willford and Behrend, 2016) found that candidates interviewed through technology got lower interviewer ratings than face-to-face candidates (d = −.41) and liked the process less (d = −.36). Studies of phone interviews point both ways: one found lower ratings by phone, another more favorable ones. The safe conclusion: use one format for everyone at a stage, and never compare a phone-screened candidate with a video-screened one on the same scale.
- One-way video feels impersonal. In a randomized study (Langer, König and Krause, 2017), people who did an asynchronous video interview found it creepier, less personal and more worrying for privacy than a videoconference, though it didn't change how attractive they found the employer.
- Design and explanation help. In data from 27,809 real applicants (Tilston and colleagues, 2023), more preparation time and the option to re-record went with better reactions, and more questions with worse ones. Explaining that everyone gets the same questions improved perceived fairness, and explaining the scheduling flexibility improved usability (Basch and Melchers, 2019).
- Automation lowers acceptance. A highly automated interview was accepted less, through lower social presence and fairness (Langer, König and Papathanasiou, 2019).
- Many people distrust AI in hiring. In Pew Research Center's survey of 11,004 US adults (December 2022), 66% said they would not want to apply to an employer that uses AI to help make hiring decisions, and 71% opposed AI making the final hiring decision. Yet 47% thought AI would do better than people at treating all applicants the same way; 15% thought worse.
When to use which
A phone screening interview still makes sense for…
- Small, scarce pools. With five applicants you want all of, a call from a person is the better first contact.
- Senior and executive roles, where the first conversation is as much about selling the job as checking the candidate.
- Logistics only. To confirm salary, location and start date, a short call or a form is enough.
- Candidates who ask for it. Offer a person as an alternative to anyone who asks, whatever format you use.
A one-way video interview fits when…
The job is on camera, and short answers to fixed questions are enough. Keep it short, give preparation time and allow a re-record: the research above links all three to better reactions. Offer a way to answer without a camera.
An AI interview fits when…
- Volume is high and recruiters can't call everyone, but you don't want to filter on a CV alone.
- You need depth in the first round: how a support agent works through a bug report, or how an engineer explains a design, with follow-up questions when the first answer is thin.
- Candidates are in many time zones or work shifts, so a live slot is hard to find.
- You want every candidate held to the same criteria, with the evidence written down for the person who decides.
How to combine them
A typical funnel
- Applications: a few knock-out questions on the form (right to work, location), not a CV keyword filter.
- First round, one format for everyone: an AI interview or a phone screen, never a mix at the same stage. Use a phone call for the candidates you are courting, and give them the same questions.
- Review: a person reads the evidence and decides who advances, with a note on why.
- Final round with your team: live, using the gaps from the first round as the starting point.
- A recruiter call before the offer: to answer questions and sell the job. That's where a human conversation pays off most.
Whatever the format, tell candidates the same questions go to everyone and they can take it when it suits them. For invitation wording, see how to invite candidates to an AI interview.
How testee runs an AI interview
- A two-way conversation. The AI interviewer asks your questions by voice, with live captions, and asks follow-ups. A candidate who can't or would rather not use a microphone chooses "I can't use a microphone or prefer to type" and types every answer instead.
- Your questions and criteria. You build the interview from blocks (conversation, quiz, coding, whiteboard, role-play, review and written response), each with the expectations a good answer covers. Candidates never see the expectations.
- Evidence, not a verdict. After the interview the analysis marks each expectation covered, covered after a prompt, partial, incorrect or not covered, with the candidate's quotes. A coverage score out of 100 is computed in code from those statuses and labeled "for review, not a hiring decision".
- The team decides. A person sets advance, hold or reject, with a required note. testee never rejects anyone on its own, and it doesn't analyze emotion, voice, face or accent.
More in how to read an AI interview report and how it works.
Summary
- Phone screen: personal, but hard to schedule, inconsistent and costly. Keep it for scarce candidates, senior roles and anyone who asks.
- One-way video interview: consistent and flexible, but no follow-ups, and every recording still needs watching.
- AI interview: consistent, flexible, with follow-ups and a report to read. Some candidates distrust it; a person must check and decide.
- In every format: one format per stage, the same job-related questions, scoring against set criteria, and an alternative for candidates who need one.
Questions and answers
What is a one-way video interview?
An asynchronous video interview: the candidate records answers on camera to questions set in advance, in their own time, and reviewers watch the recordings later. Nobody responds during it, so there are no follow-up questions.
Are one-way video interviews good?
They are consistent and easy to schedule, and candidates can take them any time. On average candidates react to them less well than to live interviews; preparation time, the option to re-record and fewer questions help. They work best when short and paired with a scorecard.
Is an AI interview the same as a one-way video interview?
No. In an AI interview the candidate has a conversation: the AI interviewer reacts to each answer and asks follow-ups. In a one-way video interview the candidate records answers to fixed questions and nothing responds.
AI interview vs human interview: which is better?
Neither in every case. An AI interview scales, asks everyone the same questions and writes down the evidence; a person builds a relationship, sells the job and judges unusual cases. Most teams use an AI interview for the first round and people for the later ones, with a person making every decision.
When is a phone screen still better?
When you have few candidates and want all of them, for senior roles where the first call sells the job, and for any candidate who asks to speak with a person.
Sources
- Blacksmith, Willford and Behrend (2016), Technology in the employment interview: a meta-analysis and future research agenda
- Langer, König and Krause (2017), Examining digital interviews for personnel selection: applicant reactions and interviewer ratings
- Basch and Melchers (2019), Fair and flexible?! Explanations can improve applicant reactions toward asynchronous video interviews
- Tilston, Krings, Roulin, Bourdage and Fetzer (2023), Reactions to asynchronous video interviews: the role of design decisions and applicant age and gender
- Langer, König and Papathanasiou (2019), Highly automated job interviews: acceptance under the influence of stakes
- Silvester, Anderson, Haddleton, Cunningham-Snell and Gibb (2000), A cross-modal comparison of telephone and face-to-face selection interviews in graduate recruitment
- Straus, Miles and Levesque (2001), The effects of videoconference, telephone, and face-to-face media on interviewer and applicant judgments in employment interviews
- Pew Research Center (2023), AI in hiring and evaluating workers: what Americans think