AI Screening Interviews Compared: Phone Bots, One-Way Video and Voice Interviews
Five first-round formats side by side: recruiter phone screens, AI phone bots, one-way video, chat screening and browser voice interviews, and when each fits.
An AI screening interview is a first-round interview run by software instead of a recruiter: the candidate answers a set of questions by phone, on video, in a chat or by voice in a web browser, and the hiring team reviews the answers before anyone moves on. The formats differ in three ways that matter: how much they ask of the candidate, whether they can ask a follow-up question, and what evidence ends up in front of the recruiter.
There are five first-round formats to choose from, counting the recruiter's own phone screen, which is still the benchmark the others get measured against. This guide describes each one, puts them side by side in a table, untangles what vendors mean by an "AI voice interview", and matches the formats to the kinds of roles you hire for.
What is an AI screening interview?
The screen is the conversation between the application and the hiring manager. Its job is narrow: confirm the hard requirements, check that the experience on the résumé is real and relevant, find out why the person is looking and what they want, and agree on the logistics, from start date to pay. A good screen answers one question for the hiring manager: is this person worth an hour of their time?
Two kinds of product get called AI screening. Résumé screeners read applications and rank them. Screening interviews talk to the candidate. Both are sold as AI candidate screening software or automated candidate screening, so check which one a vendor means. This guide is about the second kind, the tools that hold the first-round conversation.
Teams automate the screen because the volume has outgrown the phone. Gem's 2025 Recruiting Benchmarks report, a vendor report published in January 2025 on hiring data from 2021 through 2024, found that the average recruiter manages 2.7 times as many applications as three years earlier (more than 2,500), and that average time to hire rose from 33 days to 41. No recruiter can phone every plausible applicant in a pile that size quickly, and some of the people left waiting take other offers.
What are the five first-round screening formats?
1. The recruiter's phone screen
A recruiter books a time and calls the candidate. It is still the best format for judgment and for selling: a good recruiter hears hesitation, follows the interesting thread, answers any question about the team, and starts a relationship.
Its costs are all about time. Booking the call takes a round of messages, the call has to fit inside both people's working day, and the candidate has to step away from their current job to take it. Questions drift from candidate to candidate and from recruiter to recruiter, and the evidence the hiring manager gets is whatever made it into the notes.
Best for: senior and hard-to-fill roles, low volume, and any role where the first conversation is a pitch.
2. AI phone-call screening bots
An AI voice agent phones the candidate after they apply, or gives them a number to ring. It asks the screening questions out loud, records the answers, and in some products books the next interview. Some bots read a fixed script and capture the answers; others hold a looser conversation and can follow up.
The phone is familiar and needs no app, no internet connection and no camera, which suits hourly roles. The weak point is reaching people: a call from an unfamiliar number is easy to miss or ignore, and a call that arrives during a shift can't be taken. Look for retries, a call-back number and a way for the candidate to pick a time.
Best for: high-volume hourly roles where candidates live on their phones.
3. One-way video interviews
The candidate opens a link, sees each question on screen, and records an answer on camera, usually with a set time to think and a set time to answer. Reviewers watch the recordings later. Some products add AI scoring of the answers; the recording session itself doesn't change.
One-way video removes scheduling and lets a reviewer get through a stack of candidates quickly. But it asks a lot of the candidate: a camera, a quiet and presentable space, and the awkwardness of talking to a screen that never reacts. Because the questions are fixed, there is no follow-up, so a vague answer stays vague.
Best for: roles where on-camera presence is part of the job, and programs that want every candidate to answer identical prompts.
4. Chat and text screening
The candidate answers questions in a chat window on the careers site, by text message or in a messaging app, often with buttons for common answers. Chat is quick, works on a phone and suits candidates who can't talk at work. It handles availability, location, shift preferences and hard requirements well, and some chat tools book the next step too.
What chat can't give you is depth or a sense of how someone speaks. Typed answers are short, and they are easy to draft with a chatbot. Branching logic can send a candidate down a different path depending on an answer, but that is not the same as asking "what happened next?"
Best for: logistics and hard requirements at volume, where the screen is mostly yes, no and when.
5. AI voice interviews in the browser
The candidate opens a link and talks with an AI interviewer in their web browser. It is a spoken conversation, so the interviewer can follow up on a thin answer the way a recruiter would, and the candidate needs a microphone but no camera. There is nothing to schedule. The team gets a recording and a transcript, and usually a scorecard.
The costs: the candidate needs a quiet place and a device with a browser, some candidates would rather talk to a person, and the quality of the result depends heavily on the rubric behind the scoring and on whether each score shows its evidence.
Best for: roles where spoken communication and real examples matter, at a volume the team can't phone.
How do the screening formats compare side by side?
| Format | Candidate effort | Scheduling | Follow-up questions | Camera needed | Evidence the recruiter gets | Recruiter time | Best for |
|---|---|---|---|---|---|---|---|
| Recruiter phone screen | Low, but the call must fit into working hours | Booked; messages back and forth | Yes, as good as the recruiter | No | The recruiter's notes | The call, plus booking and write-up | Senior, hard-to-fill and low-volume roles |
| AI phone-call bot | Low: answer the phone or ring a number | Little: the bot rings, or the candidate does | Depends on the product | No | Recording or transcript, captured answers, sometimes a score | Reviewing results and chasing missed calls | High-volume hourly roles |
| One-way video | Higher: camera, a presentable space, answers against a timer | None, before a deadline | No; questions are fixed | Yes | A video per question; some add transcripts and AI scores | Watching the recordings | Roles where on-camera presence matters |
| Chat or text screening | Low: typing on a phone | None | Limited; usually branching on set answers | No | Structured answers and the chat log | Low: mostly checking flags | Logistics and hard requirements at volume |
| Browser voice interview | Medium: a microphone, a quiet place, a few uninterrupted minutes | None; any time | Yes | No | Recording, transcript, usually a scorecard | Reading transcripts and checking scores | Roles where speaking and real examples matter |
Two rows of that table decide most choices. "Follow-up questions" is the difference between a form and an interview: without a follow-up, a candidate who says "I handled escalations" never has to say how many, how bad or what they did. "Evidence the recruiter gets" decides how much you can check: notes from memory, a captured yes or no, or the candidate's own words with a score attached.
Is an AI voice interview the same as an AI phone screen?
Not necessarily, and vendors don't use the term consistently. "AI voice interview" and "AI voice interview platform" are used for two different things:
- A phone-based screen. The AI calls the candidate's phone, or the candidate calls a number. The conversation happens over a phone line.
- A browser-based interview. The candidate opens a link and talks with the AI on a laptop or phone browser, using the device's microphone.
Both are spoken, both can follow up, and both can produce a recording and a transcript. The differences are in reaching the candidate and in timing:
| Phone-based AI screen | Browser-based AI voice interview | |
|---|---|---|
| How the candidate is reached | A call to their number, or a number to ring | A link in an email, a text or a job posting |
| When it happens | When the call connects, or when the candidate rings in | Whenever the candidate opens the link |
| What the candidate needs | A phone | A device with a browser and a microphone |
| The usual failure | The call is missed or comes at a bad time | The link sits unopened in an inbox |
Ask every vendor one question before you compare anything else: "Does your AI phone the candidate, or does the candidate open a link?" The answer changes how you invite people, how you chase them and what the candidate experiences.
One-way video interview vs AI interview: what's the difference?
A one-way video interview is a recording session. The questions are set in advance, nothing responds to what the candidate says, and a camera is required. Where AI is involved, it is applied afterwards, to score or summarize the recordings.
A conversational AI interview is a conversation. The next question depends on the last answer, so a vague claim gets a follow-up. Voice formats need no camera, which matters to candidates who don't want to be judged on their kitchen or their face, and to you, because what you want to evaluate is what they said.
If one-way video works for you, there is no reason to drop it for roles where on-camera presence is part of the job. If you are using it because it was the only asynchronous option, look at the alternatives. Our comparison of one-way video interview software goes through them.
What should an AI pre-screening tool ask?
The same things a good recruiter asks on a first call, in the same order for everyone: logistics and hard requirements, a walk through the current role, one or two examples of the core skill, motivation, and compensation. A screen that runs about 15 minutes has room for ten or twelve questions with follow-ups, not thirty.
For a bank of questions grouped by purpose, with what each one tells you, see our 50 pre-screening interview questions. Hard requirements deserve particular care: a knockout question works best as a flag a recruiter reviews, not an automatic rejection, because candidates misread questions and licenses go by different names.
Which screening format fits which hiring?
| Hiring situation | What the first round must find out | Formats that fit |
|---|---|---|
| Hourly retail, warehouse or food service at volume | Availability, location, shifts, start date | Chat or text screening, or a phone bot |
| Call centers and customer service | How the candidate speaks, handles a difficult customer, explains | A voice format: phone bot or browser voice interview |
| Nurses and care staff | License, shifts, clinical examples, how they work with patients | Browser voice interview or chat for logistics, with licenses flagged for review |
| Professional and graduate roles at volume | Real examples, reasoning, motivation | Browser voice interview, or one-way video if on-camera presence matters |
| Staffing agencies re-screening a database | Current availability, recent experience, pay expectations | Browser voice interview or chat, sent to past candidates |
| Senior, executive and hard-to-fill roles | Fit, motivation, whether they'll move at all | The recruiter's own phone screen |
You can also combine them: chat for the yes-or-no logistics, then a conversational interview for the candidates who clear them. Whatever you choose, give candidates a way to reach a person. In Greenhouse's May 2026 survey of 1,200 US job seekers, a vendor survey, 46% wanted the option to request a human interview instead (Greenhouse, May 2026).
How do you choose AI interview screening software?
Start with the format, then test the product on your own roles. Five checks cover most of it:
- Reach. How does the candidate get the interview, and what happens when they don't respond?
- Follow-ups. Does the tool ask a follow-up when an answer is thin, or move on to the next question regardless?
- Evidence. Can you see the candidate's own words behind every score?
- Control. Does anything get rejected or advanced without a person? Knockouts should flag, not reject.
- Candidate experience. Is the candidate told it is an AI before they start, how long it takes, and what happens next?
Then run a short pilot: send the tool to real candidates on one or two roles, have two of your recruiters score 20 to 30 of the interviews with the same rubric, and compare. Our guide to the best AI interview software compares vendors, and what an AI interviewer is covers how the category works end to end. For the full vendor checklist, see how to choose AI interview software.
Where does DIO fit?
DIO, the interviewer Dio makes, is the fifth format: a structured voice interview in the browser. It does not phone anyone. The recruiter shares a link, either a private link for one candidate or the role's open link posted with the job, and the candidate takes the interview whenever it suits them. They need a microphone; a camera is never required. Most interviews take about 15 minutes, and they are in English only.
A screening round follows the arc of a recruiter's first conversation: background, what the candidate does, achievements, how they work, why they're moving, why this role, what they want, and compensation and logistics. DIO asks follow-ups when an answer is thin or vague. Knockout questions flag an answer for the recruiter's review and never end the interview. The team gets the recording, the transcript and a scorecard with quoted evidence: each competency score must quote the candidate's exact words, checked against the transcript. A person on the team makes every decision.