AI interviewers: what they are, how they work, and how to pick one for hiring

This guide is for employers: recruiters, HR leads and owners who hire the same roles again and again and want the first-round phone screen off their calendar. If you are a candidate with an AI interview coming up, our candidate FAQ explains what to expect.

Most of what ranks for “AI interviewer” today is either a practice app for job seekers or a vendor page. Neither tells a hiring team the two things that matter before you buy: whether you can check the scores, and what the law asks of you. This guide covers both, and ends with a checklist you can take into any demo.

What is an AI interviewer?

An AI interviewer is software that holds the first-round screening interview with a job candidate, the call a recruiter would otherwise make. It asks about the requirements of the role, follows up when an answer is thin, records and transcribes the conversation, and hands your team a scored report. A person on your team reads the report and decides who moves on.

Two things set it apart from the tools that came before. Older video interview platforms recorded candidates answering fixed questions and left the watching to you. Chatbots asked knockout questions by text. An AI interviewer holds a back-and-forth conversation and turns it into evidence your team can read in two minutes.

How an AI interview works in the first round

The flow looks much the same across vendors:

• You set the requirements. The tool reads your job description and pulls out the criteria. You check them, then add your own questions or knockouts.
• Candidates get an invite. By email, text, a phone call or your ATS, depending on the tool. Some vendors call candidates the minute they apply; others send a link they can open whenever they like.
• The candidate talks to the AI. Usually 10 to 20 minutes, at night or on a weekend if that suits them. A good interviewer asks one requirement at a time and probes vague answers.
• Your team reads the report. Scores per requirement, the transcript, the recording, and a ranking across everyone who finished.

The payoff is time. You stop spending a recruiter’s week on 15-minute calls that end in a no, and every applicant gets a real interview instead of a resume filter deciding for them.

Types of AI interviews: voice, video, chat and one-way

“AI interview” covers four formats that behave in different ways. The biggest split is conversational (the AI listens and asks follow-ups) versus one-way (the candidate records answers to fixed prompts).

For front-line hiring, voice usually wins. Candidates apply from their phones, many of them between shifts, and a voice interview asks for nothing more than a quiet room. Add video for the roles that need it, and check that your tool lets you choose per job.

How AI interview scoring works, and the evidence to demand

Most AI interviewers transcribe the conversation and use a language model to judge the answers against criteria. The question for you is what comes out the other end. There are two broad designs:

• One overall score. The model reads the whole interview and outputs a fit number, sometimes with a paragraph of summary. Easy to sort by, hard to check. When a hiring manager asks why someone got a 62, nobody can point to an answer.
• Criterion-by-criterion scoring. Each requirement from the job description gets its own call (met, partly met, not met) with the quote or timestamp that supports it. You can spot-check any call in seconds, and one bad answer only affects the requirement it was about.

Ask for the second design. Then, in the demo, ask to see a real report and pick a score at random. If you can’t find the sentence that justifies it within a minute, your hiring managers won’t either.

Two more things to ask about. First, what the tool measures besides the words: some products have analyzed facial expressions or tone of voice, which is harder to explain and harder to defend. Second, who sets the pass mark. Your team should choose whom to move forward, by score, by hand or both.

Where an AI interviewer fits, and where it doesn’t

It fits best where you hire the same roles over and over, with more applicants than your team can call: staffing and RPO desks, call centers and BPOs, clinics, multi-site retail and hospitality, and small businesses with no recruiter at all. In those settings the first-round questions repeat, the requirements are concrete (shift availability, certifications, years in a similar role), and speed decides who you get.

It fits less well for senior, specialized or relationship-driven hires, where the first conversation is also a pitch and a recruiter’s judgment matters from minute one. If you hire five people a year, you can make the calls yourself. And no AI interviewer should make the hiring decision. Use it to decide who your team talks to next.

Candidate experience and disclosure

Candidates judge you by the first round. A few practices separate a good AI interview from one that drives people away:

• Tell them it’s an AI. In the invite and at the start of the call. In a 2026 Greenhouse survey, 70% of candidates who had faced an AI interview said nobody told them upfront (Greenhouse).
• Let them choose the time. An on-demand link beats an unexpected call for people who are at work when they apply.
• Keep it short and about the job. Questions should map to the job description, and the interviewer should say up front how long it will take.
• Offer another route. Some candidates need an accommodation or a different format. Say how to ask for one.
• Tell them who decides. A line saying a person reviews every interview does more for trust than any amount of friendly copy.

Bias audits and the law

US law treats an AI interviewer like any other selection procedure, and a few places add their own rules. This is a summary for orientation, current as of October 26, 2026. It is not legal advice; check with your counsel for the roles and places you hire.

New York City: Local Law 144

If you use an automated employment decision tool for jobs in New York City, the law requires a bias audit of the tool within one year before use, a public summary of the results, and a notice to candidates at least 10 business days before you use it. The city’s consumer and worker protection agency has enforced it since July 5, 2023 (NYC DCWP). The audit compares scoring or selection rates across sex and race or ethnicity, and their intersections. Ask any vendor for its audit summary and the date of the audit, and note that the duty to post it and notify candidates sits with you, the employer.

Illinois: AI Video Interview Act and the Human Rights Act

The Artificial Intelligence Video Interview Act (820 ILCS 42) applies when AI analyzes video interviews for Illinois-based roles. Before the interview, you must tell the candidate that AI may analyze the video, explain how it works and what it looks at, and get their consent. You may share the videos only with people whose expertise you need to evaluate the candidate, and you must delete them within 30 days of the candidate’s request, copies included. If you keep the interview voice-only, ask your counsel whether the Act reaches you.

Separately, Public Act 103-0804 amended the Illinois Human Rights Act from January 1, 2026: employers may not use AI that has the effect of discriminating against protected classes in hiring, may not use zip codes as a proxy for them, and must notify people when they use AI in those decisions.

Federal: Title VII and the EEOC

Title VII of the Civil Rights Act covers an AI interviewer the same way it covers a written test or a structured interview. The federal Uniform Guidelines on Employee Selection Procedures set out the four-fifths rule: a selection rate for any group below 80% of the highest group’s rate counts as evidence of adverse impact under the Guidelines in most cases (29 CFR 1607.4(D)). Most bias audits, including those under Local Law 144, report impact ratios on the same logic.

Federal enforcement has shifted. In September 2025 the EEOC told its offices to close charges that rest only on disparate impact, and to notify those workers that they can take their case to court themselves (Associated Press). The statute did not change, and state and city rules still apply, so keep auditing.

How to choose an AI interviewer: a buyer checklist

Take these questions into every demo and ask for the answers in writing.

• Scoring. Does every score break down by requirement, with the quote or timestamp behind it? Can I listen to the recording and read the full transcript?
• Control. Who decides which candidates to invite and who moves forward? Can I edit the requirements and add my own questions?
• Format. Voice, video, chat or one-way? Can I choose per job? What happens when a call drops?
• Signals. Does it analyze faces, voice or tone, or only what the candidate says?
• Bias audit. Has an independent auditor tested it? When, on what data, and is the summary public?
• Disclosure. Does the interviewer say it’s an AI? Can I give candidates notice and another format where the law requires it?
• Data. How long do you keep recordings? Can candidates ask for deletion? Do you train models on our interviews?
• Languages and channels. Which languages does it interview in? Email, SMS, phone call, WhatsApp?
• ATS. Does it work with my ATS, and can it run without one?
• Price. Can I see a price before a sales call, and can I test it on a live role for free?

For how seven vendors answer these on their own websites, with prices, see our comparison of AI screening and interview software.

How VeraHire runs the first-round interview

VeraHire is our product, so here is how it answers the checklist, from what we publish on this site.

• Format. A voice conversation, in English, that the candidate takes whenever it suits them, usually in 10 to 15 minutes. Video is available on request: once we switch it on for your account, you choose which jobs require the camera. No face analysis either way.
• Invites. A link by email, or by SMS outside the US and China. You invite everyone above a resume score, the whole pool or people you pick, or let Autopilot mode send them.
• Questions. Built from the requirements in your job description, which you check before anything goes out. You can add your own questions and focus areas.
• Scoring. Each requirement marked met, partly met or not met, tied to a line in the transcript, with the recording next to it. Candidates rank as interviews finish. Your team makes the hire.
• Disclosure. The interviewer introduces itself as an AI at the start of the call.
• Bias audit. Joellie AI, an independent auditor, applied the Local Law 144 standard to our resume screening and interview analysis and found no conclusive evidence of disparate impact. It is not a certification. Read the audit summary.
• ATS. Works alongside 50+ ATS and HRIS integrations, or runs the first round with no ATS. If you don’t have one, see the free ATS.
• Price. 30 AI interviews and 300 resume screens free, no credit card. Then $7 per AI interview and $0.25 per resume screen, or plans from $49 a month billed annually. See pricing.

Where it falls short: interviews run in English only, VeraHire doesn’t place outbound phone calls or use WhatsApp, and it doesn’t book the next-round interview for you. If you need those, the comparison covers vendors that do.

Before the interview, VeraHire can also score every resume against the job, so you interview the people worth a conversation. Teams in staffing, BPO, clinics and multi-site retail and hospitality use it on the roles they fill every month. The AI interviewer product page shows a full report, or you can take a sample interview yourself, no signup.

FAQ

Legal sources checked October 26, 2026. This guide is general information, not legal advice. If something here is out of date, tell us at hello@verahire.ai.