What is an AI interview? Formats, how they work, and when to use one

A plain explanation of AI interviews for recruiting teams and candidates, with the formats side by side.

The short answer

What is an AI interview? It is a job interview, usually a first round, where software asks the questions instead of a recruiter, and then produces a summary or score for a human to review. The term covers very different products. Some run a live, two-way conversation by voice or video that follows up on each answer. Others record answers to fixed questions and score them later, or run a text chat. Candidates usually take AI interviews on their own schedule, and the hiring team gets a consistent first round for every applicant. The format you choose decides how much you actually learn.

What is an AI interview? The main types

When vendors say "AI interviews," they mean one of these. The examples are drawn from what each company publishes about its own product.

Format	What the candidate does	Does the AI react to answers?	Examples
Live AI interview (voice or avatar)	Talks with an AI interviewer in real time	Yes: follow-ups, and in some tools, difficulty changes	HireVue AI Interviewer (voice), CodeSignal AI Interviewer, Ribbon, Talview Ivy
AI interview inside a coding canvas	Codes or designs while talking with the AI	Yes	HackerRank's Chakra
One-way (asynchronous) video	Records answers to fixed, pre-set prompts	No, answers are reviewed or scored afterward	Spark Hire, Hireflix, Willo
Text chat interview	Types answers to a structured chat	Partly, depending on the tool	Sapia.ai
Scripted AI phone screen	Answers a phone call that follows the employer's script	Limited	Classet (Joy)
AI notetaker for human interviews	Talks with a human interviewer while AI takes notes	Not the AI's role	Metaview's notetaker, BrightHire

Product lines change quickly. Metaview, for example, added a voice screening agent in September 2026, and BrightHire, now owned by Zoom, sells an asynchronous AI screen alongside its notetaker. Check the current format before you assume.

What happens in an AI interview

From the candidate's side, a typical live AI interview runs like this:

Invite. The candidate gets a link by email or text, often right after applying, and starts whenever they like.
Setup and disclosure. A short check of microphone and camera, an explanation that the interviewer is an AI, and what's being assessed.
The conversation. A set of questions, usually tied to the job's key skills. In a live format, the AI asks follow-ups ("What did you do next?") and may adjust difficulty. Technical interviews can include live coding or a whiteboard.
Wrap-up. The candidate can ask questions or add anything they missed.
Review. The hiring team gets a transcript or recording, scores by skill, and highlights. A person decides who moves forward.

In a one-way video interview, step 3 is different: the candidate sees a prompt, gets a set time to think, and records an answer, with no follow-up.

Why teams use AI interviews

Speed. Every applicant can interview the day they apply, with no calendar back-and-forth. A role with 200 applicants no longer needs 50 hours of recruiter phone screens.
Consistency. Every candidate gets the same structure and rubric. US OPM research summaries link structured interviews with higher validity and less adverse impact than unstructured ones.
Coverage. You can interview more of the pool instead of filtering on resumes alone, which helps candidates whose resumes undersell them.
Candidate convenience. Candidates can interview at night or on weekends, around a current job.

Limits to be honest about

Not every candidate likes them. Some find talking to an AI awkward. Clear disclosure, a practice option and a human contact help.
Scores are only as good as the rubric. A vague rubric gives vague scores. See how AI interviews are scored.
Accessibility. Candidates with disabilities, speech differences or weak internet need an equivalent alternative.
They don't replace the final rounds. AI interviews work best as a first or second screen. Team fit, closing and complex judgment calls still belong with people.
Regulation. Using AI to screen candidates is regulated in some places. In New York City, an AI interview that drives who advances can require an annual independent bias audit and notice to candidates. In the EU, the AI Act treats AI used to evaluate candidates as high-risk, with those duties applying from December 2, 2027. See our NYC Local Law 144 guide. This is not legal advice.

Live adaptive vs one-way: why the format matters

The biggest difference between AI interviews is whether the interviewer can react. A fixed prompt like "Tell me about a hard project" gets whatever the candidate decides to say. A live interviewer can ask "What was your part in it?" or "How did you measure success?", which is exactly where strong and weak candidates separate. Adaptive interviews go a step further by making questions harder when a candidate is doing well and easier when they're struggling, so you learn where their ceiling is. Our post on adaptive questioning goes deeper, and our one-way vs live AI interview guide compares the two formats in detail.

For reference, NTRVSTA runs live AI interviews as a video avatar or audio only, adapts questions and difficulty to each answer, and covers technical and non-technical roles.

Where AI interviews fit in a hiring process

Most teams use an AI interview to replace the recruiter phone screen, or the first technical screen, not the whole loop. A typical flow for a support or sales role with 300 applicants looks like this:

Apply: 300 applications come in through the ATS.
Resume screen: knockout questions or AI resume screening narrow the pool, say to 120.
AI interview: all 120 get an invite the same day. Perhaps 90 complete it.
Recruiter review: a recruiter reads the score reports and advances 15.
Human rounds: the hiring manager and team interview the 15 and make the decision.

For engineering roles, the AI interview often sits where a take-home or a first technical call used to, with the human loop focused on system design, depth and team fit. Either way, the AI round should be the same for everyone in that role, so the review step compares like with like.

How to choose an AI interview tool

Take the interview yourself as a candidate, for one technical and one non-technical role.
Read a real score report. Can you trace each score to something the candidate said?
Ask for the bias audit if you hire in New York City, and check its date and auditor.
Work out the cost per completed interview at your volume. Our AI interview pricing guide shows how.
Check fit: your ATS, languages, accommodations and how candidates are told it's an AI.

When you're ready to shortlist, our list of the best AI interview software compares the main options. If you're a candidate, read how to prepare for an AI interview.

Questions

Is an AI interview a real interview?

Yes, in the sense that your answers are recorded and used in the hiring decision. It is usually the first round. Treat it with the same preparation as a recruiter call, and expect a human to review the results before any final decision.

How long does an AI interview take?

It varies by tool and role. Fountain describes its frontline AI interviews as about 5 minutes, while Interviewer.AI's plans allow 20 or 30 minute interviews. Technical interviews with coding or system design usually take longer than a short screening call.

Does the AI decide whether I get the job?

It shouldn't. Vendors including HireVue and CodeSignal say humans make the final decision, and good practice keeps a person accountable for every decision, with access to the evidence behind each score.

What's the difference between an AI interview and a one-way video interview?

A one-way video interview records your answers to fixed prompts and doesn't respond. A live AI interview is a two-way conversation in which the AI follows up on what you say, and sometimes changes difficulty as you go.

The fastest way to understand an AI interview is to take one. Start a sample interview, or book a 20-minute demo to see the hiring team's view.

Sources

NTRVSTA