What is an AI interview?

How an AI interview differs from other hiring tools

The label gets applied loosely. A one-way video recorder that plays the same five prompts to every candidate is often marketed as an AI interview, but it does not adapt follow-ups based on what someone actually said. A chatbot that collects scheduling preferences is also not an interview in the hiring sense — there is no structured evaluation of job-relevant answers.

A genuine AI interview sits closer to what a trained recruiter does on a first-round phone screen: it holds a conversation, probes incomplete answers, and applies a consistent rubric across candidates. The difference is scale and standardisation — the same question logic and scoring criteria can run at 2am on a Sunday for candidate 400 just as it did for candidate 1 on Monday morning.

That consistency is the main reason teams adopt it for high-volume roles where human phone screens become a bottleneck. It is not a replacement for every interview stage — final hiring decisions, culture fit, and specialised technical depth still typically belong to humans — but it can absorb the repetitive structured screening that otherwise burns recruiter hours.

What happens during a typical AI interview

1. Candidate joins on their own schedule

   Most AI interviews are asynchronous: the candidate opens a link on a phone or laptop when convenient, without coordinating a calendar slot with a recruiter.

2. The system conducts a structured conversation

   Questions follow a role-specific rubric. Strong implementations adapt follow-ups based on prior answers rather than reading a fixed script verbatim.

3. Responses are captured and scored

   Spoken answers are transcribed, evaluated against predefined criteria, and summarised for the hiring team — ideally with evidence pointing to specific moments in the transcript.

4. Results feed the existing pipeline

   Scores and transcripts typically export to an ATS or recruiter dashboard so the next human stage starts with structured signal, not a blank resume.

Where AI interviews work well — and where they do not

They tend to work best for roles with clear, repeatable screening criteria: spoken communication in a contact center, shift availability in a warehouse, credential status for a nursing hire, or basic sales pitch clarity for a telecaller. The questions have right and wrong shapes even if answers are open-ended.

They work less well as the sole gate for roles requiring deep specialist judgment — senior engineering architecture, executive leadership, or clinical skills assessment — where a live expert interviewer or hands-on exercise remains necessary. Teams that treat AI screening as one stage in a funnel, not the entire funnel, get better outcomes and fewer candidate complaints.

Legal and fairness considerations matter regardless of role. Any automated screening that affects who advances should be monitored for disparate impact, documented for audit, and governed under applicable hiring laws — topics covered in our glossary entries on adverse impact, EEOC compliance, and regional rules like NYC Local Law 144.

Voice vs text AI interviews

• Voice interviews test what phone-heavy jobs actually require — clarity, pace, tone under pressure — in a medium closer to the job itself.
• Text interviews reduce accent and connectivity friction but measure written fluency, which may not correlate with spoken performance in voice roles.
• Hybrid flows (voice with text fallback) appear in markets with uneven mobile connectivity, common in bulk hiring across India.
• Latency and transcription accuracy materially affect candidate experience; poor ASR on accented speech produces bad follow-up questions that feel like the system is not listening.

Frequently asked

Related